ON RESUMPTIVE PRONOUNS IN SLAVIC

VRINDA SUBHALAXMI CHIDAMBARAM

A DISSERTATION

PRESENTED TO THE FACULTY

OF PRINCETON UNIVERSITY

IN CANDIDACY FOR THE DEGREE

OF DOCTOR OF PHILOSOPHY

RECOMMENDED FOR ACCEPTANCE

BY THE DEPARTMENT OF

SLAVIC LANGUAGES AND LITERATURES

AND THE PROGRAM IN LINGUISTICS

Adviser: Edwin S. Williams

June 2013

ii

© Copyright by Vrinda S. Chidambaram, 2013. All rights reserved.

iii

Abstract

"On Resumptive Pronouns in Slavic" is a first step toward a generalized theory of

pronouns that accounts for the behavior of both ordinary and resumptive pronouns. I

adopt the definition of resumption proposed by Boeckx (2003) who analyzes resumptive

pronouns as pronouns stranded by the movement of an NP sister. In contrast to Boeckx's

theory, I argue that every definite pronoun enters the syntax as a D0 sister to its NP or DP

referent.

DPMAXIMAL

D DPINTERNAL

her

D NP

the linguist

The central argument of the dissertation rests on this proposed universal pronominal

structure, called the stacked DP. I suggest that the single parametric difference resulting

in resumptive vs. non-resumptive type languages is the ability of the internal DP to raise

independently of the pronoun. In other words, any language in which the internal DP of a

stacked DP is independently mobile (i.e. it can move without pied-piping the pronoun)

will contain resumptive pronouns.

This analysis of the syntactic structure of pronouns extends to all instances of

pronouns, including those that do not occur in a resumptive context. A natural question to

follow from this is what becomes of the NP/DP sister of the pronoun in contexts not

involving any resumption, as in the following sentence:

(i) I met her yesterday.

iv

I propose that the NP/DP referent is frequently deleted by a PF operation, Pronominal

Associate Deletion (PAD):

If D0 is a personal pronoun and XP is a sister to D

0, then XP is deleted.

If the XP referent raises during narrow syntax, the conditions for PAD will not be met at

PF. As a result, both the XP and the pronoun will be pronounced.

In addition to this PF rule, I propose an LF condition to account for the

interpretation of pronouns, the Pronominal Reference Condition (PRC):

In DPMAX

whose D0 is a pronoun, interpret D

0 as co-referential with the NP also

dominated by DPMAX

.

These two operations, in combination with the stacked DP internal structure of pronouns,

provide the means to explain a wide array of phenomena relating to the occurrence of

resumptive pronouns, including Macedonian clitic doubling, Hebrew interrogative wh-

resumption, optional resumption in B/C/S and Slovak, and resumption in Slovene

superlative clauses.

v

Acknowledgments

I have a great many people to thank for helping me to research and write on the

topic of resumptive pronouns in Slavic and an even greater number of people to thank

who helped me maintain my sanity as I researched and wrote. There is a good deal of

overlap between the two groups. I am forever grateful to you all.

First and foremost, I would like to express my sincerest gratitude to Edwin

Williams, who took me on as a student and had confidence in me. Without his

encouragement, I would never have been able to write this. His incisive questions

exposed many of the problems and lacunae in my analysis, forced me to confront

complex details directly, and thus made my arguments far stronger than they would have

been otherwise. He never gave me answers, but always suggested the path that would

lead me to them, thus giving me the means to arrive at the solutions independently. I am

particularly grateful for the genuine interest that Edwin Williams took in my ideas. Even

more meaningful than his invaluable critique of my analysis was his faith in my abilities

as a linguist.

I would also like to thank E. Wayles Browne and Steven Franks, whose guidance

led me to Slavic linguistics and to the study of South Slavic languages. Many of the ideas

that I present here were formed by reading their work. And they have always been willing

to read what I have written, to comment on it, to offer their suggestions and support. I am

deeply thankful that I have had such wonderful mentors over the years, and I aspire to be

like them, not only because they are brilliant linguists but also because they are true

friends and genuinely kind people.

vi

I would like to thank Leonard H. Babby, who brought me to Princeton and taught

me such a tremendous amount while I was there. I will always remember his seminars

with the greatest fondness. His insistence on clarity and straightforwardness improved my

writing exponentially. I hope that whenever I write, I will always hear his voice in my

ear, reminding me that the simplest phrasing is always the best.

I cannot express enough gratitude to my many friends and colleagues whom I

constantly bothered with questions about translations, glosses, and grammaticality:

Pavlos Avlamis, Muamera Bregović, Márton Dornbach, Branislav Kovalčík, Mária

Miskarik, Jozef A. Müller, Phoevos Panagiotidis, Biljana and Žane Temova, Ognen

Vangelov, and Zuzka Vatralová. And I would like to offer a special thanks to all my

wonderful Slovenian linguists and language informants: Luca Barone, Alja Ferme, Marc

Greenberg, Marko Hladnik, Peter Jurgec, Lanko Marušič, Janez Orešnik, Amanda

Saksida, Adrian Stegovec, Nejc Štukl, Gašper Tkačik, and Sašo Živanović.

I thank my many friends and family, who have been so loving and supportive. I

don't know where I would be without you all. I particularly want to thank my choirs:

K'vali (Ljubljana, Slovenia), Niavi (Washington, D.C.), and Dedebi/Gaumarjos

(Princeton, NJ) – I might have finished my thesis sooner if I hadn't been singing so much,

but then again, I might not have made it through graduate school at all if I hadn't. I also

want to thank all of the people who were essential to my life during my graduate studies:

Zaur Agayev, Cori Anderson, Margaret Beissinger, Anna Berman, Theodor Brasoveanu,

Fran Carrol, Lindsay Ceballos, Geoff Cebula, Katherine Drake, Christine Dunbar, SiYi

Feng, Kate Fischer, Maggie Goldman, John Graham, Olga Hasty, Carl Linich, Jamie

McGavran, Chris Minkowski, Matthew Moss, Serguei Oushakine, Mark Pettus, Tim

vii

Portice, Leeore Schnairsohn, Stas Shvabrin, Brian Street, Jason Strudler, Elizabeth

Timberlake, Sarah Valentine, Michael Wachtel, Emily Wang, Cornel West, Tessie and

Tim West, Andrew James Young, and the D-bar (I consider it a friend).

I am grateful to my parents, Raju and Shobha Chidambaram. As the daughter of a

mathematician and a neurologist, I suppose it is no wonder I turned out to be a linguist. I

cannot thank them enough for all the love, support, encouragement, and everything else

they have given me. I would also like to thank my sister, Sharada. As the sister of an

attorney, it is certainly no wonder that I developed (purely for self-preservation) some

skills in arguing. She has always been and always will be the person that I admire the

most and that I most wish I could be like. Above all, I would like to thank my husband,

Jozef Müller, who encouraged me onward, who was invariably patient and kind when I

was being intolerable, and who always put my needs before his own. I would never have

been able to write this without his help.

viii

Table of Contents

ABSTRACT iii

ACKNOWLEDGMENTS v

LIST OF ABBREVIATIONS 1

INTRODUCTION 2

Overview 2

Framework 3

Syntactic Operations 4

Conditions on Syntax 5

CHAPTER 1: ISSUES PERTAINING TO THE PHENOMENON OF

RESUMPTION 6

1.1. Defining Canonical Resumption 6

1.1.1. Relative Clause Gaps 7

1.1.2. The Analysis of Relative Clauses 10

1.1.2.1. The Raising Analysis: Verngaud (1974) and Kayne (1994) 11

1.1.2.2. The Adjunction Analysis: Chomsky (1977) 20

1.1.2.3. A Hybrid Model - Two Structures of Relative Clauses:

Sauerland (1998, 2003) 25

1.1.2.4. A Summary of Relative Clause Analyses 30

1.1.3. A Pronoun and Not a Gap 31

1.1.4. The Definition of a Canonical Resumptive Pronoun 36

1.2. The Properties of Pronouns 36

1.2.1. Binding Theory and Resumption 38

1.2.1.1. Binding Condition B as a Constraint on Interpretation 39

ix

1.2.1.2. The Internal Structure of Pronouns 41

1.3. Previous Analyses of Resumptive Pronouns 43

1.3.1. Some Early Analyses 43

1.3.2. Boeckx (2003) and the Stranding Analysis 44

CHAPTER 2: A NEW ANALYSIS OF CANONICAL RESUMPTIVE

PRONOUNS 48

2.1. Requirements on the Structure 48

2.1.1. The Status of Clitic Pronouns 50

2.1.2. Checking Definiteness on the Pronoun 52

2.1.2.1. The Definiteness of Relative Clause internal DPs:

Evidence from Macedonian Clitic Doubling 54

2.1.3. The Interpretation of Resumptive Pronouns 57

2.2. A New Analysis 63

2.2.1. The Structure of the Relative Clause

and the Structure of the Pronoun 63

2.2.1.1. The Matching Model 63

2.2.1.2. The Stacked-DP Pronoun Structure 64

2.2.1.3. Languages with and without Canonical Resumption 68

2.2.1.4. Island Effects 68

2.2.2. Derivations of Canonical Resumption 71

2.2.2.1. Bosnian/Croatian/Serbian:

Feature-Checking and Licensing Co-Reference 71

2.2.2.2. Hebrew: A Case of Optional Resumption 74

2.2.2.3. Slovene: A Case of "Clitics Only" Resumption 77

2.2.2.4. A Universal Condition on Resumptive Pronouns 81

x

2.2.3. Overt C0 Relatives versus Wh-Relatives in Slavic 82

2.2.3.1. What Happens When an Overt Wh-word Merges? 83

2.2.3.2. What Happens When an Overt Wh-word Raises? 85

2.2.3.3. Getting the Facts 85

2.2.3.4. The Structure of Wh-Relatives versus Overt C0 Relatives 86

2.2.3.4.1. The Girl with Whom I Danced 88

2.2.3.4.2. The Girl that I Danced with Her 89

2.2.4 Split Antecedents and Canonical Resumption 90

2.3. Summary 91

2.3.1. Two Predictions 93

2.3.1.1. Hungarian 94

2.3.1.2. Greek 95

CHAPTER 3: A BRIEF EXAMINATION OF TWO INSTANCES OF NON-

CANONICAL RESUMPTION 97

3.1. Hebrew Interrogatives 97

3.1.1. An Aside on English 100

3.2. Macedonian Object Clitic Doubling 102

3.3. Summary 107

CHAPTER 4: OPTIONAL CANONICAL RESUMPTION IN B/C/S

AND SLOVAK 109

4.1. Introduction 109

4.2. Morphological Matching Analysis (Gračanin-Yuksek 2010) 111

4.3. Slovak 116

4.3.1. Slovak Data Corrsponding to B/C/S Data 116

xi

4.3.2. Slovak Free Relatives 118

4.4. Optional Resumption Under the Stacked-DP Analysis 120

4.4.1. Optional Non-Oblique Resumptive Pronouns 120

4.4.2. Obligatory Oblique Resumptive Pronouns 123

4.5. Summary 124

CHAPTER 5: A FEATURE MISMATCH MYSTERY IN SLOVENE 126

5.1. A Brief Typology of Slovene Relative Clauses 126

5.1.1. Kateri-type Relatives (which-relatives) 127

5.1.2. Free Relatives and Co-Relatives 128

5.1.2.1. The Morphology of Relative Wh-pronouns 128

5.1.2.2. The Function of Relative Wh-pronouns 130

5.1.3. Ki-type Relatives (that-relatives) 132

5.2. Slovene Superlative Clauses (kar-phrases) 134

5.2.1. Ki-type Relatives versus Kar-phrases 135

5.2.2. A Descriptive Overview of the Syntax of Kar-phrases 138

5.2.3. A Descriptive Overview of the Semantics of Kar-phrases 141

5.3. Comparatives in Slovene 144

5.3.1. Comparatives and Kar-phrases 147

5.3.2. The Morphology of Slovene Comparatives and Superlatives 149

5.3.3. A Formal Analysis of Comparatives and Kar-Phrases 151

5.4. A Formal Analysis of Superlatives and Kar-phrases 154

5.5. The Resumptive Pronoun in Kar-phrases 156

5.6. The Interaction with Mass Nouns 157

xii

5.7. Languages Other than Slovene 158

5.8. Ki-relatives with Superlative Modified DPs 162

5.9. Summary 166

CONCLUSION 169

APPENDIX A 172

BIBLIOGRAPHY 175

1

LIST OF ABBREVIATIONS

1st = first person

2nd = second person

3rd = third person

ABL = ablative case

ACC = accusative case

ADV = adverb

ANIM = animate

AOR = aorist tense

AUX = auxiliary verb

COMP = complementizer

compA = comparative adjective

DAT = dative case

DEF = definite

DU = dual number

FEM = feminine gender

FOC = focus

FUT = future aspect/tense

GEN = genitive case

IMP = imperfect aspect

INAN = inanimate

INST = instrumental case

LOC - locative case

MASC = masculine gender

NEG = negation

NEUT = neuter gender

NOM = nominative case

OP = silent wh-operator

PAST = past tense

PL = plural number

PRES = present tense

PREP = preposition

PRON = pronominal

RC = relative clause

RP = resumptive pronoun

SG = singular number

supA = superlative adjective

2

INTRODUCTION

Overview

This thesis is an investigation of resumptive pronouns. A resumptive pronoun has

three fundamental properties: it is a pronoun, it appears at the origin site of a movement

operation, and it takes the moved item as its antecedent. The typical environment in

which a resumptive pronoun is found is the relative clause. While there has been great

deal of research devoted to the syntax and semantics of relative clauses, pronouns, and

those pronominal elements within relative clauses known as resumptive pronouns, these

studies have not been exhaustive. This thesis, too, is not intended to be an exhaustive

survey of all the phenomena linked to resumption, but will rather have two specific goals.

The first goal is to develop an analysis for what I will call canonical resumption, which

is, as will be established in detail in this chapter, the occurrence of a pronoun in the site

of relativization of a relative clause. In the Bosnian/Croatian/Serbian (B/C/S) example

below, the relative clause is bracketed and the resumptive pronoun (which occurs in a

clitic pronominal form) is indicated in bold font.

Bosnian/Croatian/Serbian

(1) Čovjek [RC što sam ga vidio juče ]

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC. sawMASC.SG.PAST. yesterday

is here

je ovdje.

'The man that I saw (him) yesterday is here.'

Canonical resumption occurs in a number of unrelated languages, such as Hebrew, a

Semitic Language (Borer 1984, Shlonsky 1992), Vata, a Kru language (Koopman and

Sportiche 1982), and Bosnian/Croatian/Serbian, Slavic languages (Browne 1986). Cross-

3

linguistically, canonical resumption involves a pronominal element (either a long form

pronoun or a clitic pronoun1) within a relative clause. Because the two salient

components of canonical resumption are: 1) the pronoun and 2) the relative clause, it is

necessary to conduct a thorough review of the current theories of pronouns and relative

clauses before presenting an analysis of resumptive pronouns.

Having constructed an analysis of canonical resumption, I will then examine

instances of non-canonical resumption that occur in the South Slavic languages. In

particular I will be looking at Slovene, B/C/S, and Macedonian. These three languages all

exhibit canonical resumption, but additionally, in each of them we find instances of

resumptive pronouns appearing in contexts unlike those found in (1); that is, they do not

fit the pattern of canonical resumption. All three languages present data that conflict with

existing theories of resumption. The second goal of this thesis will be to develop an

analysis of each of these non-canonical resumptive constructions that coheres with the

broader theory of resumption.

Framework

The framework I adopt for this study is the Principles and Parameters framework.

This style of analysis was initially developed by Noam Chomsky and was motivated by

the notion of a universal generative grammar; that is, a non-learned grammatical

computational system, part of which is innate, and which, by means of various

algorithms, may generate novel utterances. Since the advent of this analytical model, it

has been the generative linguist's main goal to discover the various algorithms at work

and identify which of them are universal and which are language-specific.

1 Many languages contain two types of pronominal forms: the long form, which is typically used when the

pronoun is focussed, and the clitic form, which is used everywhere else.

4

Syntactic Operations

The basic machinery within the current P&P framework are the three operations:

Merge, Copy, and Delete (Chomsky 2008). Merge can be characterized as having two

main functions: to enter syntactic items into the derivation and to extend the derivation by

re-Merging these syntactic items as higher nodes in the derivation. While I do not dispute

the notion that there is only one operation at work (i.e. Merge), in the ensuing chapters I

will refer to re-Merge operations as movement and will use the term Merge to refer only

to First Merge. I do this only for the sake of simplicity.

The operation Copy occurs whenever a syntactic item is moved. The moved item

leaves behind an exact copy of itself in the position which it occupied, and the highest

copy (i.e. merged at the highest position within the derivation with respect to the other

copies) is pronounced (Pesetksy 1998). This particular operation will be crucial in the

forthcoming chapters, because it replaces the previous Government & Binding concept of

traces. In much of the early literature on resumption, resumptive pronouns were

considered overt traces of moved elements. Since traces have been replaced, given the

Copy Theory of movement, this way of describing resumption is invalidated. The

analyses which I will put forward will be consistent with the Copy theory of movement.

The function of the Delete operation is to ensure that features illegible to PF or LF

do not proceed to those interfaces. This is assuming a Y-model of syntax in which the

syntactic derivation proceeds (perhaps cyclically) to Spell-Out, at which the derivation

diverges into two distinct paths: Logical Form (where the derivation continues to the

conceptual-intensional component) and Phonological Form (where the derivation

continues to the sensory-motor component).

5

These three basic operations are applied to satisfy syntactic requirements. For

example, an item will move in order to check its features (i.e. Case or φ-features) against

those of another item occurring higher in the derivation. I adopt the feature-checking

theory proposed by Simpson (2000) in which a syntactic item may check it features with

those of another syntactic item when the structural relationship between the two is either

a Spec-Head or a Head-Complement relation.

Finally, I assume that abstract case on a DP is a feature like any other and must be

checked by means of a specific configurational relationship to a case-checking Head.

Case may be checked when the DP is either a Specifier or a Complement of the case

checking head.

Conditions on Syntax

Beyond the constraints on the three syntactic operations, there are also more

general requirements on the syntax; conditions that must be met in order for the

derivation to converge. For example, the condition on the interpretation of anaphors

requires that an anaphor have a certain structural relation to its antecedent (namely, the

antecedent must c-command the anaphor within domain D (Chomsky and Lasnik 1977))

in order for it to be interpretable. In this thesis, I will adopt several conditions formulated

by others (i.e. Pronounce Highest Copy (Pesetsky 1998, Franks 1998)) and will introduce

these conditions as they become relevant to the analysis. In the following chapters my

aim is to demonstrate how the three syntactic operations outlined here as well as specific

conditions on derivations will result in the appearance of both canonical and non-

canonical resumptive pronouns.

6

CHAPTER 1

ISSUES PERTAINING TO THE PHENOMENON OF RESUMPTION

1.1. Defining Canonical Resumption

Resumption (that is, the appearance of a pronoun in the site of a moved syntactic

item) occurs in a variety of contexts (which will be addressed in later chapters), but

canonical resumption is the appearance of a resumptive pronoun in a relative clause.

While they are not found in English2, resumptive pronouns are a common feature of

many of the world's languages and in languages that are unrelated to one another. In

every language in this diverse linguistic group; that is, in languages that exhibit any sort

of resumption, the one environment in which a resumptive pronoun will certainly occur is

within a relative clause. Put differently, languages that exhibit resumption will have

resumption in relative clauses3. It is for this reason that I call such instances of

resumption canonical resumption.

2There are infrequent occurrences even in English of pronouns that appear to be resumptive. They are

primarily used as a last resort repair strategy to mend expressions that could not otherwise be parsed. The

status of these pronouns has been discussed in Sells (1984), Boeckx (2003), and recently Clemens, Morgan,

Polinsky, and Xiang, in their 2012 study Listening to Resumptives: An Auditory Experiment. All of the

above have identified the pronouns occasionally produced in English islands (known as intrusive pronouns

(Sells 1984)) as distinct from resumptive pronouns. This distinction is based on several empirical facts,

notably that there is absolutely no reliably grammatical occurrence of intrusive pronouns in English. While

some speakers may consider them to ameliorate some island effects at times, just as many others do not.

Additionally, according to the data from the experiment conducted by Polinsky et al., most speakers find

them ungrammatical across the board. That is to say, they are not, in fact, part of the English grammar. 3Spanish typically has no canonical resumptives pronouns, but it does display clitic-doubling. This would,

at first blush, seem to be a counter-example to my claim. However, clitic doubles only occur in Spanish

when they double a DP embedded in a PP. It is also the case that a resumptive pronoun will appear when

the site of relativization occurs within a PP. Consider the following data:

Yo la ví a Ana.

INOM.SG. herACC.SG.FEM. saw1st.SG.PAST. toPREP Ana

'I saw Ana.'

La mujer a la que ví era Ana.

The womanNOM.SG.FEM. toPREP herACC.SG.FEM. that saw1st.SG.PAST. was Ana

'The woman that I saw (her) was Ana.'

A proper investigation of these data would extend beyond the scope of the current topic, but they illustrate

that the generalization holds for Romance languages as well. The precise mechanisms that only allow

7

1.1.1. Relative Clause Gaps

Before addressing the phenomenon of canonical resumption itself, we must first

examine the environment in which it arises, meaning that we must first carefully consider

the properties of relative clauses. Consider sentence (2):

(2) Sharada watched the movie [RC that I mentioned].

The bracketed phrase in (2) is a relative clause; it is a clausal modifier of movie. This N0

modified by the relative clause is known as the relative head. And it is precisely the role

of the relative head that will prove crucial, because there is no general consensus as to its

precise syntactic relationship to the relative clause (e.g. while some linguists, notably

Vergnaud (1974) and Kayne (1994), believe that the relative head is first merged into a

position within the relative clause, others, such as Chomsky (1977), believe that it is

merged outside of the relative clause. The various analyses will be discussed in detail in

section 1.1.2).

If sentence (2) contained no other elements than those that appear at PF, it would

constitute a serious deviation from standard generative principles. If we consider the

semantic roles assigned by the predicates, the relative head, movie, appears at first to be

assigned two thematic roles at once: both as the thing that I mentioned and as the thing

that Sharada watched. Given the θ-criterion, this would present a problem. In fact,

however, the θ-role assigned by watched is assigned to the entire DP the movie that I

mentioned, while the θ-role assigned by mentioned is assigned only to N movie. So the

semantics, in terms of theta-assignment, do not pose a problem.

resumption following a preposition are unclear.

8

However, given only the PF-legible elements, the syntax is highly problematic.

The verb mention obligatorily takes a direct object complement (* Jozef mentioned.) and

obligatorily assigns its complement accusative case. In sentence (2), mentioned does not

have a direct object complement in the canonical English post-verbal position. Instead,

there is a gap4. And in fact, such a gap exists in all English relative clauses - it can act as

the subject, direct object, indirect object, or as any other grammatical function within the

relative clause.

Subject Gap:

(3) Raju chased the squirrel [RC that ______ bit my finger].

In sentence (3), the gap occurs in the highest subject position of the relative clause. While

the interpretation is unproblematic (i.e. it was the squirrel that bit my finger), the subject

of bit is not identified in the DP-Spec-TP of the relative clause.

Indirect Object Gap:

(4) Shobha instructed the woman [RC that she gave the jasmine plant to _______ ] to

water it daily.

The gap in (4) occurs as the indirect object within the relative clause. Again, the

interpretation is uncontroversial - the N woman and the indirect object of the relative

clause have the same denotation, but the indirect object is unpronounced.

Adjunct Gap:

(5) The couch [RC that Jozef fell asleep on ______ ] had a broken leg.

4A gap is defined as a phrase that is elided on identity with a higher phrase.

9

We see in (5) that the relative clause gap can also occur within an adjunct phrase. All of

the argument positions required by the verbal predicates are filled, but within the relative

clause, there is an adjunct prepositional phrase, headed by P0 on, whose complement is

unpronounced. And although the PP is merely an adjunct, it becomes crucial to the

grammaticality of the sentence because it contains the gap.

(5) a. * The couch [ that Jozef fell asleep ] had a broken leg.

In sentence (5a), all of the argument positions are filled, and there are thus no violations

of the θ-criterion nor of the case assignment criteria: The matrix predicate have requires

two arguments, to which it assigns its two θ-roles: possessor (assigned to the subject

couch) and theme (assigned to broken leg). In addition, the matrix predicate assigns

accusative case to the DP a broken leg. The embedded predicate fall asleep requires only

one experiencer θ-role which it assigns to its subject, Jozef. The subjects of both clauses

are assigned structural nominative case. Thus, case and thematic criteria are not the

source of the ungrammticality of (5a). The problem with (5a) is rather that the

relationship between the embedded clause and the matrix clause is entirely undefined.

The gap in relative clauses is crucial precisely because it determines that relationship5.

The English relative clause is not unique in containing a gap: all PF structures in

which a modifier and its modified object are non-contiguous generally contain a gap.

This also applies to the syntax of predicate adjectives. Under most contemporary analyses

(since Stowell 1981), the predicate adjective is first merged with the modified noun,

5In Turkish, which has neither relative pronouns nor complementizers, the only way to distinguish between

a relative clause and a subordinate clause is that the former contains a gap. Many thanks to Leonard Babby

for bringing this to my attention.

10

producing a small clause, as in (6). Subsequently the noun raises out of the small clause

giving the order found in (7).

(6) [ is [ Sharada clever]]

(7) [ Sharada is [ clever]]

The gap in (7) is a consequence of movement6. Similarly, the gap found in relative

clauses is also considered to be a consequence of movement. The mechanics of this

movement, however, are still being debated. Various theories have been developed to

describe and explain the presence of a gap in relative clauses (see Vergnaud 1974,

Chomsky 1977, Kayne 1992, Sauerland 1998). In other words, there are a number of

different ways to derive the structure of a relative clause, but all of them involve a

movement operation. In the following section, I examine the current theories of relative

clause syntax.

1.1.2. The Analysis of Relative Clauses

Because the derivation of canonical resumption is inalienable from the derivation

of relative clauses, it is imperative that, before launching into a discussion of canonical

resumption, we look carefully at the structure of relative clauses. The difficulty in

examining relative clauses, however, is that there is no consensus as to their structure. In

fact, there have been two competing theories since the late 1970s and neither has surfaced

as the clear victor. Therefore, in this section I will examine both theories of relative

clause structure.

6Babby (2009) points out that a similar movement operation is also responsible for the distribution of short

form adjectives in Russian.

11

The first that I will discuss is based on an analysis initially proposed by Jean-

Roger Vergnaud in his 1974 MIT dissertation, French Relative Clauses. Vergnaud's

analysis is that the relative head originates at the site of relativization within the relative

clause and subsequently raises to a higher position outside the relative clause where it is

then pronounced. This theory was later adopted and elaborated upon by Richard Kayne

(1994).

The second theory is based on Chomsky's (1977) analysis of relative clause

structure. Chomsky's proposal was that the relative head is base-generated in a position

external to the relative clause. Within the relative clause there is a silent operator (Op)

which orginates at the site of relativization (i.e. the position of the gap in the relative

clause) and ultimately raises to a higher position. The Op bears a relationship to the

relative head, namely, the two are co-indexed and thus co-referential.

In the following section, I will look at how each of these theories handles simple

relative clauses that don't involve resumption. I will look at their treatment of both

relative clauses headed by an overt C0 as well as wh-relatives. Finally, I will also consider

a more recent proposal (Uli Sauerland 1998, 2003) that combines the Vergnaud-Kayne

and the Chomsky models into a hybrid analysis of relative clauses.

1.1.2.1. The Raising Analysis: Vergnaud 1974 and Kayne 19947

The critical assumption guiding the Vergnaud-Kayne syntactic analysis of relative

clauses is that the relative head initially occupies a position within the relative clause,

7What I describe herafter will be the Vergnaud-Kayne model. At the time Vergnaud first proposed this

analysis of relative clauses, a great deal of terminology, such as CP and merge, did not yet exist. While the

theory that I summarize here is certainly the one initially conceived by Vergnaud and subsequently

advanced by Kayne, I will be using updated terminology, in keeping with current models of syntax. That is,

I will discuss the same theory but in terms of operations such as merge, lexical categories such as C, and

excluding elements such as traces.

12

namely the site of relativization. That is, in examples (2)-(5), repeated here, the relative

head originates in the position of the gap. (Moved items are indicated in italics).

(2) Sharada watched the [ movie [ that I mentioned movie ]].

(3) Raju chased the [ squirrel [ that squirrel bit my finger]].

(4) Shobha instructed the [ woman [ that she gave the jasmine plant to woman ]] to water

it daily.

(5) The [ couch [ that Jozef fell asleep on couch]] had a broken leg.

In these sentences, the relative head is first merged into a position within the relative CP.

It subsequently raises to the Specifier of the relative CP, leaving behind a copy in its

original and in each of its successive positions.

(2) Sharada watched the [CP moviei [C' that I mentioned moviei]].8

In (2), there is a relative clause CP embedded within the matrix clause. In the initial stage,

as syntactic items are being merged into the derivation, the NP movie is merged as the

complement of V mentioned. So, movie is first merged within the relative clause. As the

derivation proceeds, the subject of the relative clause, I, and the C0, that, are merged. At

that point in the derivation, the NP movie raises to its eventual Spell-Out position, as

Spec-CP of the relative clause.

Kayne's (1994) analysis of wh-relatives is not drastically different from the

analysis he outlines for relative clauses with overt C, insofar as both involve a movement

8Given Chomsky's Inclusiveness Principle, any indexation of lexical or functional items in the syntax is

impossible. The use of indices here is only to mark members of a chain for the sake of clarity. They are not

intended to be syntactically real or LF legible elements.

13

of the relative head from the site of relativization to Spec-CP. Wh-relatives, however,

require two distinct movement operations.

(8) Sharada watched the [CP [whichP moviej [which' which moviej]i] [C' I mentioned [which movie]i]].

The phrase which movie is initially merged as the direct object of V mentioned. The first

movement operation raises the entire phrase which movie into Spec-CP. The second

movement raises NP movie to a position higher than which; according to Kayne (1994),

the Spec-whichP. This final movement results in the desired PF word order, however it is

not a clearly motivated operation. In fact, there are several problems with the derivation

shown in (8) as a whole.

The two most readily apparent problems involve the role of the Determiner.

Kayne (1994) identifies which as the head of a whichP. This proposal undermines an

important generalization: which is a wh-word that never directly precedes a determiner in

English. That is, which occurs in complementary distribution with other determiners, a

fact indicating that which most likely belongs to the lexical category D. Furthermore, the

determiner the and the NP do not at any point in the derivation form a constituent; rather

D obligatorily takes a CP argument. This appears to be a unique case of D merging to CP.

In all other cases, this is impossible, as illustrated in examples (9) - (11) below.

(9) * [DP the [CP John left]].

(10) * [DP the [CP what did John say]] annoyed me.

14

It appears that D is only able to merge to a relative CP. But even this restriction is not

narrow enough to prevent over-generation. Consider (11):

(11) * [DP the [CP what John said]] annoyed me.

The D0 the is merged to a relative CP and nevertheless the resulting sentence is

ungrammatical. The reason for this is that the relative clause in (11) is a free relative (i.e.

a relative clause containing a wh-relativizer and missing an over relative head). The

restriction, then, seems to be that a Determiner can only merge with a relative CP

containing an overt relative head. This amended form of Kayne's proposal will generate

sentences such as (12), which are in fact attested in some dialects of British English.

(12) [DP the [CP thing what John said]] annoyed me.

Nevertheless, there is no clear reason for determiners only to take relative CPs

(and only headed-relative CPs at that) as complements and for any other combination of

D+CP to be ungrammatical. Leaving this issue open, there is also a question regarding

the relationship that obtains between the determiner and the relative head. Bianchi (2000)

points out that under Kayne's analysis, the determiner and the relative head never form a

constituent. This poses a problem for languages that exhibit agreement between

determiners and NPs. Bianchi proposes a solution to this problem, noting the Spec-Head

relationship that holds between D and whichP at the end of the derivation; that is, the

whichP is in Spec-CP, and an agreement relationship obtains between D and Spec-CP.

Bianchi (2000) also addresses the problem of motivating the movement of NP

from its initial position within the whichP into Spec-whichP. She proposes that the

15

determiner has a strong [+N] feature, which drives the movement of the relative head into

the specifier position, thereby allowing D to check its [+N] feature via the Spec-Head

relation. The presence of a [+N] feature on D is also indepentently motivated, given that

determiners most often require an NP argument and that there are many languages whose

determiners show morphological agreement with their complement NPs.

While there are many possible problems with Kayne's analysis of wh-relatives,

the most complex issue that arises applies to both the Vergnaud-Kayne analysis of overt-

C-relatives as well Kayne's analysis of wh-relatives. This is the issue of Case assignment.

Consider the following sentences:

(13) The [CP songi [C' that my friends wrote songi]] was amazing.

(14) The [CP [whichP songj [which' which songj]i] my friends wrote [whichPwhich song]i] was

amazing.

In both (13) and (14), the phrase containing the relative head N moves from its original

position as the direct object of V wrote, an accusative case marked position, into Spec-

CP, which is not a case-marked position. It would be expected, then, that the relative head

would remain marked for accusative case. This is not, however, what the data show. In

languages which exhibit morphological case, the case of the relative head is determined

by its grammatical function within the matrix clause and not by its relationship to the

relative clause predicate9. Consider the following data from Czech, a West Slavic

language with rich inflectional morphology for nouns.

9In some languages (e.g. German), there are instances of "case matching effects" in free relatives, in which

the wh-word must satisfy the case-checking requirements of both the relative clause and the matrix clause,

as in the following example from Groos and van Riemsdijk (1979):

16

Czech:

(15) Mé kamarádky napsali písničku.

myNOM.PL.FEM. friendsNOM.PL.FEM. wrotePL.PAST. songACC.SG.FEM.

'My friends wrote a song'

In (15) we see that the verb write in Czech assigns accusative case to its complement. The

N, which in the nominative would be písnička, becomes písničku in the accusative; that

is, it takes the feminine accusative singular -u ending. If, according to the Vergnaud-

Kayne model, the relative head begins in an accusative case marked position and

subsequently raises to Spec-CP, then we would expect that the relative head be spelled

out as an accusative case marked noun. This is not, in fact, what happens, as illustrated in

(16):

(16) Písnička, co napsali mé kamarádky, byla super.

songNOM.SG.FEM. that wrotePL.PAST. my friendsNOM.PL.FEM. wasSG.FEM.PAST. amazing

'The song that my friends wrote was amazing.'

Ich vertraue wem du vertraust.

I trust whomDAT.SG.MASC. you trust.

'I trust whomever you trust.'

Here, the same predicate (vertrau-) occurs in both clauses, and its dative case checking properties are

satified by the wh-word. In the next example, however, a case conflict arises between the the matrix clause

and relative clause predicates:

Ich vertraue *wem / *wen du mir empfiehlst.

I trust whomDAT.SG.MASC. / whomACC.SG.MASC. you to-me recommend

'I trust whomever you recommend to me.'

The wh-word must check both dative case (with V vertraue) and accusative case (with the relative clause

internal predicate empfiehlst). This results in an unresovlable case conflict for German. In an ordinary

relative clause (not free relative or co-relative), the sentence would obligatorily contain two distinct case-

marked elements, as shown in the following example:

Ich vertraue dem Lehrer den du mir empfiehlst.

I trust the-teacherDAT.SG.MASC. himACC.SG.MASC. you to-me recommend.

'I trust the teacher you recommend (him) to me.'

17

In (16) we see that the relative head is in the nominative case, since it is the subject of the

matrix clause. This means that although the relative head merges into the derivation in an

accusative case-marked A-position and then raises into a caseless A'-position, the form of

the relative head is not the accusative písničku, but rather the nominative písnička.

(17) Písnička, kterou napsali mé kamarádky, byla

songNOM.SG.FEM. whichACC.SG.FEM. wrotePL.PAST. my friendsNOM.PL.FEM. wasSG.FEM.PAST.

super.

amazing

'The song that my friends wrote was amazing.'

The same surprising facts hold for example (17) containing a wh-relative. And even more

surprising, given the analysis, is that which and song do not appear in the same case:

while písnička ('song') is nominative, kterou ('which') is accusative.

Valentina Bianchi (2000) offers a possible solution to this problem. First she

assumes that the whichP is, in fact, a DP and that which is a determiner. She then

proposes that Case is a feature of D0 and that an NP will simply agree with its governing

D0 in case.

18

Diagram 1

TP

TP

DP

T VP

D CP

Ønom V AP

DP CP was amazing

NPj DPi C TP

songNOM Ø

D NPj DP TP

whichACC songACC

my friends T VP

V DPi

wroteacc

D NP

whichACC songACC

Looking at diagram (1), which is the derivation of (17)10

, we see that the DP containing

the relative head is merged with the V wrote, which assigns accusative. The entire DP

then raises to Spec-CP, and then the NP song raises into Specifier of DP. Given Bianchi's

analysis, each NP must check its case feature with its nearest governing D. At first

glance, this may seem to work: song checks its case with its nearest governing D, a null

determiner with a nominative case feature, and the V wrote checks accusative case on the

determiner which. There is, however, a serious flaw: if an NP checks its case with its

nearest governing D, then the case feature of song should be accusative and this feature

should be checked by D which while song is in its first merge position. Then, when NP

song raises to Spec-DP, it will also check nominative case with the null determiner. This

is a clear case conflict - the NP is assigned two distinct abstract cases. One way to resolve

10

In the derivations of Czech examples, I use the translated English words instead of the original Czech

words, but the structures reflect Czech.

19

this is to apply Pesetsky's (2007) analysis of case conflict in Russian quantified nominals.

Pesetsky (2007) claims that higher affixal projections suppress case features acquired

lower in the derivation, and thus the higher case is pronounced. Given his analysis, the

NP in fact does receive case in both positions but the morphological case that is realized

on the NP is the one that it acquires highest in the derivation, thus the nominative.

The advantage to this movement-based account is that it provides an automatic

explanation for certain reconstruction effects. Consider the following sentence:

(18) The [CP [NP picture of himself] [RC that John took [NP picture of himself ]]] hangs on

the office wall.

At PF, the anaphor himself appears in a position higher than its antecedent John -

a configuration that would be ruled out under any formulation of Binding Condition A.

However, given the movement operation in which the relative head picture of himself

originates as the complement of took, one could formulate Condition A to take into

account the original position of the anaphor with respect to its antecedent. In its original

position, himself is c-commanded by John within the minimal CP, and thus there is no

violation of Condition A.

Nevertheless, in order to accept the Vergnaud-Kayne model of overt-C-relatives

and Kayne's model of wh-relatives, one must also accept a number of other tangentially

related proposals, such as Bianchi's case-assignment analysis and Pesetsky's Pronounce

Higher Case proposal.

20

1.1.2.2. The Adjunction Analysis: Chomsky 1977

Chomsky's (1977)11

account of relative clauses, like the Vergnaud-Kayne account,

involves a movement operation, but the two accounts are distinguished by the nature of

the moved item. Under the Vergnaud-Kayne analysis, the moved item is the relative head.

Under Chomsky's analysis, it is either a wh-word or a silent wh-Operator that is moved.

The wh-element (either having a PF legible set of features or a PF illegible operator) is

first merged into the site of relativization (that is, the same position into which, according

to the Vergnaud-Kayne model, the relative head would be merged) and subsequently

raises to Spec-CP of the relative clause. This immediately offers a more parsimonious

account of the case phenomena. Consider again the Czech example (17), derived here

according to Chomsky's (1977) analysis:

11

In 1977, Chomsky was working under a very different set of assumptions than those held by proponents

of standard generative grammar today. While the analysis he presented in 1977 is still in widespread use,

the terms needed to describe the analysis have changed. Just as I did for the Vergnaud-Kayne model, I will

discuss Chomsky's analysis using a modified set of assumptions (that is, the set of assumptions mentioned

previously and those regularly adopted by linguists working today in a P&P framework: Copy Theory,

existence of CP, existence of DP, lack of traces and indices, etc.).

21

Diagram 2

TP

DP TP

D NP T VP

Ø

N CP V AP

song was amazing

DPi CP

which ACC

C TP

Ø

DP TP

my friendsNOM T VP

V DPi

wrote whichACC

The relative head, song (písnička), receives its case in the same way that the nominative

subject of any clause receives its case - it occurs in Spec-TP and checks its nominative

case feature with a non-defective, finite T. There is no case conflict and thus no need to

invoke Pesetsky's (2007) case-conflict resolution strategy. Furthermore, the process of

relativization is reduced to a single movement operation for wh-relatives. And this

operation is, in fact, identical for both wh-relatives and overt-C relatives. The following is

the derivation of sentence (16), a Czech sentence containing overt C, according to the

Chomsky model:

22

Diagram 3

TP

DP TP

D NP T VP

Ø

N CP V AP

song was amazing

DPi CP

wh-OPACC

C TP

that

DP TP

my friendsNOM T VP

V DPi

wrote wh-OPACC

Comparing the two Chomsky-analysis-based derivations shown above, one will notice

that the only two differences are these: in the second derivation 1) C contains the PF-

legible item that and 2) the wh-word is PF-illegible.

The Chomsky model of relative clauses, in addition to lacking case-assignment

complexities and divergent structural analyses of overt C versus wh-relatives, is able also

to account for data that are beyond the explanatory reach of the Vergnaud-Kayne model.

For example, Chomsky's analysis allows for a specific structural variance that will result

in the difference between restrictive and non-restrictive relative clauses.

23

Diagram 4

Restrictive:

DP

D NP

the

N CP

professor

DP CP

who

who teaches philosophy

In this restrictive relative clause, the relative CP is a sister to the relative head N, thus

forming an NP. The determiner is a sister to NP, thus taking scope over the entire NP

containing the relative clause.

Diagram 5

Non-Restrictive:

DP

DP CP

D NP DP CP

the professor who

who teaches philosophy

In the non-restrictive relative clause, the relative clause CP would be a sister to a DP

comprised of the determiner and the NP professor. The determiner in Diagram 5 only has

the NP professor within its scope and not the relative clause.

24

Kayne (1994) proposes that the difference between restrictive and non-restrictive

relative clauses is due to a difference in their LF structures; the TP within the relative

clause raises to Spec-DP at LF only in non-restrictive relatives. While this does provide

some account of the semantic differences between restrictives and non-restrictives, if we

take a strict view of the Y-model, Kayne's theory is problematic. There are not only

semantic differences but also phonological differences between restrictive and non-

restrictive relative clauses; namely, there is a prosodic break that occurs before and after

the relative clause in non-restrictives that does not occur in restrictives.

While Chomsky's analysis is preferable to the Vergnaud-Kayne model for several

reasons (e.g. it does not involve case conflict and it can account for the differences

between restrictive and non-restrictive relatives), it too has drawbacks. One clear

disadvantage of Chomsky's analysis is that it, unlike the Vergnaud-Kayne analysis, does

not provide a natural account of reconstruction effects. Since the relative head is first

merged outside of the relative clause, there is no obvious reason why there would not be

any apparent violation of Condition A in sentences like (18), where according to

Chomsky's analysis, the anaphor would be merged into a position higher than its

antecedent.

In addition, the D which in a wh-relative has no nominal complement at any point

in the derivation. In every other context in which we find it, D which is part of a D-linked

wh-phrase (e.g. which song). Thus, having the determiner which appear without an NP

complement only occurs in the syntactic environment of a relative clause. This leaves us

with two possibilities: either this determiner which (i.e. the one occurring in relative

clauses) must be syncretic with but lexically distinct from other instances of D0 which or

25

a special rule applies to D which in the context of a relative clause, allowing it to appear

without an NP complement.

A related but more serious problem with Chomsky's model is that the relationship

between the relative clause and the relative head is not entirely clear. When Chomsky

developed this model, indices were commonly used as syntactic markers of co-reference.

Given the Inclusiveness Principle, indices can no longer function in that capacity. Instead,

the only occasion in which we can say two items are automatically co-referential is when

one is a copy of the other and thus they bear identical features. In the Vergnaud-Kayne

model, the relative head moves from the site of relativization (leaving a copy of itself in

that position) into the Spec-CP position where it is spelled out, thus the co-reference

between the pronounced item and its lower copy is automatic. In Chomsky's model, in

which the relative head is merged in a position external to the relative clause, there is a

wh-operator that must be, by some means, referentially linked to the relative head. It is

not at all clear how that referential link is manifested in the syntax.

1.1.2.3. A Hybrid Model - Two Structures of Relative Clauses

(Sauerland 1998, Sauerland 2003)

In his 1998 MIT dissertation, Uli Sauerland proposed an analysis of relative

clauses in which there is both an internal head (as in the Vergnaud-Kayne model) and

external head (as in the Chomsky model). He proposes that the internal head “contains a

phonlogically deleted version of the external head. Implicit in this proposal is that the

external head and the internal head must match at the level of LF” (Sauerland 1998: 79).

In a later version of the proposal, Sauerland (2003) outlines two distinct models of

relative clauses that he suggests co-exist; that is, both analyses are valid and both must be

26

available in the syntax, but they apply to different data. The first model he describes is a

modified version of the Vergnaud-Kayne raising analysis. The only differences between

that analysis and Sauerland's is that Sauerland proposes that 1) the NP movement of the

relative head raises it to a position external to the relative clause and 2) there is an

operator in the DP containing the relative head, even when there is an overt C. This

second property of Sauerland's model is similar to what we find in Chomsky's proposal.

Diagram 6 (Raising):

TP

DP TP

D NP was amazing

Ø

NPj CP

songNOM

DPi CP

D NPj C TP

(OPACC) songACC that

DP TP

my friends T VP

V DPi

wrote

D NP

(OPACC) songACC

This raising analysis suffers from some of the same problems as the Vergnaud-Kayne

raising analysis: there is case conflict on the relative head and there is no way to

distinguish restrictive from non-restrictive relatives. Sauerland argues, however, that this

raising structure must be available in the syntax, because there are relative clauses in

27

which the head can only be interpreted internally. Sauerland applies the idiom chunk test

as a diagnostic to show that reconstruction effects12

are possible in a relative clause:

(19) The call that the referee made was disputed.

The crucial phrase here, make a call, is understood as an idiom, in the sense of judging a

play during a sporting event, and does not take on a literal meaning, even though the

idiom is non-contiguous. This provides evidence of a raising analysis that allows the

relative head to reconstruct to its previous position and be interpreted internally (and only

internally) to the relative clause.

Sauerland also argues, however, that the raising structure cannot be the only

possible structure available for relative clauses. He offers data showing that, contrary to

what we would expect if we applied the raising analysis, we do not find Condition C

reconstruction effects in relative clauses.

(20) a. Which is the picture of Johni that hei likes?

b. * Which picture of Johni does hei like? (Sauerland 2003)

(21) a. The pictures of Marsdeni which hei displays prominently are generally the

attractive ones.

b. * Which pictures of Marsdeni does hei display prominently? (Safir 1998)

In both (20) and (21) above, the (b) examples show interrogative wh-movement of a

phrase containing an R-expression. The resulting sentences are deviant because they

12

Reconstruction effects are generally given as evidence that a movement operation has occurred in a

sentence. If the conditions on syntax are not met at the level of PF but are satisfied at an earlier stage in the

derivation and the sentence is therefore grammatical, then the conditions on syntax have been met via

reconstruction (i.e. the conditions are satisfied by the lower copy of a syntactic item).

28

show Condition C reconstruction effects; that is, the chains formed by the movement of

the wh-phrase cause a Condition C violation.

On the other hand, both of the (a) sentences are grammatical. If we note that the

relative head in both sets of sentences contains an R-expression and if we assume the

raising analysis, in which the relative head moves from the site of relativization into a

position external to the relative clause, it is surprising that there are no Condition C

reconstruction effects in either of the (a) sentences. We would expect the (a) sentences

and the (b) sentences to have the same factors determining grammaticality. Since that is

not the case, Sauerland proposes the matching model of relative clauses, which has the

following structure for Czech sentence (16):

Diagram 7

TP

DP TP

D NP was amazing

Ø

N CP

songNOM

DPi CP

D NP C TP

OPwh songACC that

DP TP

my friends T VP

V DPi

wrote

D NP

OPwh songACC

29

In this derivation, the relative head, song, is first merged with a phonetically null wh-

operator determiner. This DP raises into Spec-CP of the relative clause. Once there, an

operation called Relative Deletion applies:

Relative Deletion (Sauerland 2003: 31)

In matching relatives the internal head must not be pronounced. Furthermore, the external

head must be the antecedent of the internal head.

This operation is analogous to the operation of Comparative Deletion (Bresnan 1973,

1975), in which an item is deleted based on identity with an item higher in the derivation.

Sauerland does not however spell out the details of this operation. One can surmise,

however, that the internal head , once it raises to Spec-CP, is deleted when it is interpeted

as co-referential with the external head; that is any element in Spec-CP that takes the

external head as an antecedent is deleted. This deletion is obligatory13

.

This hybrid account, which allows for two different models of relative clauses,

solves many of the problems that occur in each of the previous two analyses, taken in

isolation. In the matching model, there is no case conflict and there are two distinct

possible structures for restrictive and non-restrictive relatives. Furthermore in wh-

relatives, the D which takes the relative head as its NP complement. The relationship

between the relative clause and the relative head is clear: the pronounced relative head is

the antecedent of the deleted internal head, which has moved from a position within the

relative clause to Spec-CP. However, it leaves open an important question: under what

13

Leonard Babby (p.c. 2013) mentions a stylistically marked sentence that is perhaps grammatical to some

speakers:

I found a pencil, which pencil was lost by John.

In this sentence, the Relative Deletion rule does not appear to apply. While I have found no speakers who

accept this sentence as grammatical, it is nevertheless possible that in some dialects, Relative Deletion is

optional.

30

conditions does the raising structure appear and under what conditions does the matching

structure appear? This question is left open for further research.

1.1.2.4. Summary of Relative Clause Analyses

I have considered two competing analyses for the structure of relative clauses: the

Vergnaud-Kayne analysis and the Chomsky analysis, as presented in sections 1.1.2.1 and

1.1.2.2. Each of these analyses offers some insight into the relationship between a relative

clause and the matrix clause in which it appears. Both are, however, problematic for a

variety of reasons: either the case phenomena are not clearly derived or the relationship

of the relative head to the relative clause is not clearly represented. The third analysis

presented, Sauerland's matching model, offers the possibility that there actually exist two

available structures for relative clauses in the syntax. He bases this proposal upon data

reflecting reconstruction effects. The two structures are both, in some respects, hybrids of

the Vergnaud-Kayne and Chomsky models. While they capture many of the same

structural dependencies as the previous analyses, they avoid some of the basic problems.

For this reason, I will adopt Sauerland's analysis of relative clauses.

The analysis of relative clauses is critical to the study of resumptive pronouns for

the following reason: based on how one derives a relative clause, one automatically

arrives at some notion of where a resumptive pronoun will fit into the derivation. That is

to say, the structure of a relative clause must contain a gap; we have seen that in every

one of the analyses presented. But depending on the analysis, the gap could be 1) an

unpronounced copy left from movement of the relative head (Vergnaud-Kayne model), 2)

a silent wh-operator (Chomsky model), or it could be both of these together (Sauerland

31

model). In whichever relative clause model one adopts, the resumptive pronoun will bear

a relationship to that gap.

1.1.3. A Pronoun and Not a Gap

In English and many other languages, the gap in a relative clause is quite plainly

just that - nothing is pronounced but there is some semantic and syntactic connection

between the gap in the relative clause and another term in the derivation. That is, while

nothing surfaces in PF at the site of relativization, there is a syntactic and interpretive

element within the relative clause that associates the relative clause with the matrix

clause.

In other languages, however, there is a pronounced element occupying this

position: a pronominal form. This is the canonical resumptive pronoun. Consider

sentence (1), repeated below, and its English translation.

Bosnian/Croatian/Serbian

(1) Čovjek [RC što sam ga vidio juče ]

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC. sawMASC.SG.PAST. yesterday

je ovdje.

is here

'The man that I saw (him) yesterday is here.'

The translation given for sentence (1) is literal; in fact, if the pronoun in parentheses were

actually pronounced, the English sentence would be ungrammatical:

(22) * That man that I saw him yesterday is here.

32

In English relative clauses, no PF-legible element, pronominal or otherwise, can occupy

the site of relativization. On the other hand, the resumptive pronoun is obligatory in the

B/C/S example. Put another way, the difference between these two languages is that

B/C/S exhibits canonical resumption while English does not.

In order to form a basic, descriptive definition of canonical resumption, it is

useful to look at another example and compare the two.

Bosnian/Croatian/Serbian

(23) Pričao sam sa devojkom, što si *(joj)

spokeSG.MASC. AUX1st.SG. with girlINST.SG.FEM. that AUX2nd.SG. herDAT.SG.FEM.

pomagao juče.

helpedSG.MASC. yesterday

'I spoke with the girl that you helped (her) yesterday.'

Comparing examples (1) and (23), one can immediately see some similarties: both are bi-

clausal, containing a matrix clause as well as a relative clause. Both relative clauses begin

with the word što, a morphologically invariable complementizer meaning that, and both

relative clauses obligatorily include a resumptive pronoun instead of a gap. The salient

difference between the two is the form of the pronoun. In (1) the pronoun is ga, the

singular, masculine, animate accusative pronoun. In (23), the pronoun is joj, the singular,

feminine, dative pronoun. The gender difference between the pronouns of the two relative

clauses reflects the gender difference between the relative heads of the two sentences.

The case difference reflects the difference between their case-checking heads.

In sentence (1), the relative head is čovjek, meaning man. The lexically inherent

features, or φ-features, on this noun are valued [ +masculine, +singular, +animate]. The

33

case feature, a feature which is not lexically inherent but rather structurally determined, is

nominative, since čovjek is the subject of the matrix clause. If we look at the resumptive

pronoun in (1), we see that its φ-features have identical values to those of the relative

head. That is, the pronoun ga ('him') is [+masculine, +singular, +animate]. Its case,

however, is distinct from the case of the relative head. The pronoun, too, receives

structural case but it checks accusative case with the embedded predicate vidio ('saw').

These same basic facts hold for sentence (23); only the specific values of the

features are different from those found in (1). The relative head is devojkom, meaning

girl. The φ-features of devojkom are valued as [+feminine, +singular, +animate14

]. The

case feature on devojkom is instrumental, checked by the preposition sa. The pronoun joj

shares the same φ-feature values as the relative head, but differs from the relative head in

its case feature. The case on the pronoun is dative, as it is the direct object of the lexical

dative case checking predicate pomagao.

The pronoun in the relative clause not only matches the relative head in φ-

features, but in fact, it must match the relative head in reference15

.

Bosnian/Croatian/Serbian

(24) * Čovjeki [RC što sam gaj vidio včera ]

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC. sawMASC.SG.PAST. yesterday

je ovdje.

is here

'The man that I saw (him) yesterday is here.'

14

Although animacy does not have any morphological consequence on either feminine or neuter nouns, I

will nonetheless assume that it is part of the feature complex of every B/C/S noun. 15

The pronoun ga and the N čovjek, as well as the entire NP čovjek što sam ga vidio, must have the same

denotation. They cannot refer to distinct individuals.

34

If we take the pronoun ga and the relative head čovjek, to have disjoint reference,

the resulting sentence is deviant. In order for the sentence to be grammatical, the two

must bear the same reference. Given this fact (i.e. they are obligatorily co-referential), we

can refer to the relative head as the antecedent of the resumptive pronoun.

These data reveal the fundamental syntactic function of the resumptive pronoun:

it is an element that matches the features and reference of the relative head but receives

its case based on its grammatical function within the relative clause.

Recall the data in (5) and (5a), repeated below:

(5) The couch [RC that Jozef fell asleep on ______ ] had a broken leg.

(5) a. * The couch [that Jozef fell asleep] had a broken leg.

These data illustrate that the gap in a relative clause determines the relationship between

the relative clause and the higher clause in which it is embedded (as shown in section

1.1.1). Now, let us look at a translation of (5) in Slovak, a language with canonical

resumption.

Slovak

(25) Pohovka, [RC čo na nej Jozef zaspal ] mala

couchNOM.SG.FEM. that on itLOC.SG.FEM. JozefNOM.SG.MASC. fell-asleep had

zlomenú nohu.

broken leg

'The couch that Jozef fell asleep on (it) had a broken leg.'

35

The interpretive function of the resumptive pronoun in Slovak is identical to the function

of the gap in an English relative clause; it determines the relationship of the relative

clause to the higher clause in which it is embedded.

Zaenen, Engdahl, and Malin (1981) show, using a coordination test, that a relative

clause gap and a canonical resumptive pronoun are of the same syntactic category. They

refer to an observation dating back to Ross (1967) that in coordinate structures, if an

element in one conjunct has undergone a movement operation, then that same operation

must have applied to the other conjuncts as well. Williams (1978) formalizes this

observation into a set of conditions on across-the-board rule application. One of these

conditions is that the moved elements in each conjunct must be of the same syntactic

category. Zaenen, Engdahl, and Malin offer data from Swedish, showing extraction from

conjoined relative clauses, in which one relative clause contains a gap and the other

contains a resumptive pronoun:

(26) Där borta går en mani som jag ofta träfar ___i men inte minns vad

There goes a man that I often meet ___ but don't remember what

hani heter.

he is-called

'There goes a man that I often meet but don't remember what he is called.'

(Zaenen, Engdahl, and Malin (1981))

Thus, the syntactic category of the gap and the resumptive pronoun (as well as their

interpretation) must be identical. The resumptive pronoun must therefore be part of either

an NP or a DP constituent.

36

1.1.4. The Definition of a Canonical Resumptive Pronoun16

A Canonical Resumptive Pronoun is a pronoun:

1. that occurs inside a relative clause headed by an indeclinable C0,

2. whose antecedent is the relative head,

3. whose case is determined by its grammatical function within the relative clause,

4. whose function is to determine the relationship between the relative clause and the

higher clause,

5. whose syntactic category is either N or D17

.

1.2. The Properties of Pronouns

An important generalization about resumption is that, in all languages exhibiting

resumption, the resumptive element takes the form of a pronoun. Thus, the following is

universal:

If: 1) z is an R-expression with a fully valued set of φ features,

2) x is a lexical pronoun whose φ-features are valued identically to those

of z, and

3) y is a pronounced lexical term in the structure [z[RC...y...]] in which z

is the antecedent of y,

Then: x and y are syncretic.

16

Note that English intrusive pronouns (mentioned in footnote 2) are not considered resumptive pronouns,

since they do not occur regularly, they are never considered grammatical, and they do not exhibit many of

the defining properties of resumptive pronouns (e.g. they do occur in highest subject position). 17

The question of whether the resumptive pronoun is an N0 or a D

0 will be discussed in section 1.4. There

do not appear to be any non-pronominal elements that regularly appear as resumptive pronouns. Consider

the following example from Slovak, a language that does contain canonical resumptive pronouns:

Obdobie čo Ivan strávil *vtedy na Slovensku

The-time that Ivan spent then in Slovakia

'the time that Ivan spent in Slovakia'

The sentence is made ungrammatical by the presence of vtedy, which is meant to refer to obdobje. An

adverb cannot be used in a resumptive function.

37

In other words, resumptive pronouns are universally syncretic with ordinary pronouns.

Because this is a true linguistic universal, it cannot legitimately be dismissed as mere

coincidence. The fact that resumptive pronouns and ordinary pronouns invariably have

the same phonetic forms is not fortuitous. This notion is elegantly formulated by

McCloskey:

A fundamental question, which has not often been explicitly addressed, but which

lies behind much of the discussion is why resumptive elements have the form that

they do. That is, resumptive pronouns simply are (formally) pronouns. I know of

no report of a language that uses a morphologically or lexically distinct series of

pronouns in the resumptive function. If we take this observation to be revealing,

there can be no syntactic feature which distinguishes resumptive pronouns from

ordinary pronouns, and any appeal to such a feature must be construed as, at best

an indication of the limits of understanding. (McCloskey 2006: 97)18

Unless there emerges concrete evidence indicating otherwise, we must assume that

resumptive pronouns are not lexically or syntactically different from ordinary pronouns.

We therefore cannot distinguish them based upon what they are but rather where they

appear. That is, the term resumptive refers not to a type of pronoun but rather to a context

in which a pronoun may occur. Therefore, if a theory of ordinary pronouns is unable to

account also for resumptive pronouns, then that theory is not powerful enough.

Furthermore, any assertion that the ordinary pronoun and the resumptive pronoun are

lexically or syntactically distinct motivated solely by a commitment to maintain theory-

internal consistency, while "at best an indication of the limits of understanding"

(McCloskey 2006), is at worst a case of data tampering to suit an existing theory19

.

18

This quotation came to my attention by its appearance in Asudeh (2011), who also mentions a similar

passage in McCloskey 2002: 192. 19

It is important to note that PF is certainly not always indicative of syntactic structure; that is, there are

many operators and processes in the syntax that are not observable at PF. But any assertion that the

38

The conditions on binding are of particular interest when one assumes an identity

relation between resumptive and ordinary pronouns. All pronouns are dependent semantic

variables whose reference is determined by another nominal element (or elements, in the

case of split antecedents) either within the sentence or elsewhere in the discourse or

whose reference is determined by means of deixis. As stated by Panagiotidis (2002) "it

seems to be an absolute universal that pronouns refer but do not denote." However,

beyond this fundamental fact, pronouns vary across languages in terms of the contexts in

which they appear (e.g. in pro-drop languages, pronouns in subject position can be elided

at PF) and how their interpretations are constrained (i.e. their binding conditions). The

particular binding constraint relevant to the topic of resumptive pronouns is Condition

B20

. The reason that this is particluarly relevant is that no existing formulation of

Condition B can account for the interpretation or binding properties of resumptive

pronouns. Concretely, pronouns in a resumptive function are obligatorily coreferential

with a proximate antecedent, whereas pronouns in other functions are not.

1.2.1. Binding Theory and Resumption

The deepest concern for anyone working on anaphora within the current

Principles and Parameters framework is that we no longer have indices at our diposal,

which were previously the pillar of defining conditions on binding. Given fairly recent

postulates concerning conditions on representations and derivations, the original

Government and Binding solutions to problems of anaphoric reference (that is, binding),

syntactic structure differs from the PF must be motivated by evidence beyond theory-internal requirements.

If we claim that theory-internal arguments are in and of themselves sufficient for supporting claims about

syntax, then nothing prevents us from proposing that all PF is accidental, thereby rendering all data

irrelevant. In essence, while it is certainly acceptable to posit things that are not observable, the moment we

begin to ignore the things that are observable, we can no longer call what we are doing science. 20

Condition B: A pronoun must not be bound within its binding domain.

39

become impossible - the theoretical mechanisms previously used to solve those puzzles

have become obsolete; the Inclusiveness Condition (Chomsky 1995:228), as will be

discussed in greater detail below, rules those mechanisms out. And although G&B

binding theory was not able to account for the full array of anaphoric data21

, Conditions

A, B, and C (a basic definition of these under G&B can be found in footnote 21), covered

a great deal of empirical ground. The task, then, has been to construct a set of conditions

that is equally powerful as the one formulated under G&B but whose formal mechanics

comply with the Inclusiveness Condition.

1.2.1.1. Binding Condition B as a Constraint on Interpretation

(Chomsky and Lasnik 1993, Chomsky 1993)

In 1993 Noam Chomsky and Howard Lasnik produced a re-analysis of the

Binding Theory. One of the results of the paper was that they developed a version of

Binding Theory that was compliant with a condition later introduced by Chomsky (1995),

21

There is attested data from languages that are, in a sense, "immune" to binding condition violations, as

formulated in G&B; that is :

A. an anaphor must be bound in domain D

B. a pronoun must not be bound in domain D

C. an R-expression must not be bound

The following are examples taken from Siewerska (2004) of violations of these conditions:

Condition A violation, long-distance binding of an anaphor in Gujarati (Mistry 2000: 353):

Raaji Kišorj kamiTimaa pot-annei/j nimše em lakhe che

Raj Kishor committee self-ACC will-appoint thus write AUX

'Raj writes that Kishor will appoint himself (either Raj or Kishor) to the committee'

This Gujarati sentence is ambiguous, because the anaphor may refer to either Raj or Kishor.

Condition B violation, binding of anaphor and pronoun in the same domain in Hausa (Newman 2000:524):

Tàlá tā gan tà / ga kântà à madûbîn

Tala past:3SGF see her / see herself in mirror

'Tala saw herself in the mirror.'

Although the verb takes different forms depending on whether the pronoun or the anaphor is used,

there is no reason why the pronoun should be considered to be in a different domain from the

anaphor.

Condition C violation, long-distance binding of an R-expression in Malayalam (Jayaseelam 2000:162)

Raaman paRañña Siita Raaman-e sneehikkunnu ennǝ

Raaman said Sita Raaman-ACC loves that

'Raaman said that Sita loves himRaaman.'

While there have been various theories (e.g. Fiengo and May's (1994) proposal of Vehicle Change

to rescue condition C violations) that help to account for some of the data observed, they are not

comprehensive theories (i.e. they cannot account for the full array of data).

40

namely, the Inclusiveness Condition (where N is the numeration, and π and λ are the

phonetic and logical levels of computation, respectively):

Any structure formed by the computation (in particular π and λ) is constituted of

elements already present in the lexical items already selected for N; no new

objects are added in the course of the computation apart from rearrangements of

lexical properties. (Chomsky 1995:228)

One inevitable result of this condition is that indices, which were once a crucial element

in the definition of binding, are barred. The analysis of binding proposed by Chomsky

and Lasnik (1993) does not include indices, and as pointed out by Freidin (1997), is

"derivational in nature". It offers an alternative to conditions on binding and instead

propose a series of interpretive procedures. These procedures are re-iterated and

supported by Freidin (1997):

(a) If α is an anaphor, interpret it as coreferential with a c-commanding phrase in D.

(b) If α is a pronoun, interpret it as disjoint from every c-commanding phrase in D.

(c) If α is an r-expression, interpret it as disjoint from every c-commanding phrase.

As Freidin notes, these procedures alone are insufficient, as they will not account for the

ungrammaticality of (27):

(27) * Mary expects John to like herself. (Freidin 1997)

Given only the interpretive procedures outlined above, there is no reason that herself and

John should not corefer. Freidin, therefore, proposes an addendum to the procedures,

namely that the anaphor must be interpreted with a c-commanding antecedent within

domain D. Because the φ-features of herself and John do not match, specifically the

41

gender feature, there is no possible antecedent for herself. Therefore, the interpretive

procedure cannot be completed, and the sentence is rendered uninterpretable.

Furthermore, Freidin observes that while the interpretive procedure (b) will

account for one interpretation of sentence (28), given a CP domain, it does not offer any

explanation as to the source of the interpretation in which Mary and she are coreferential.

(28) Mary thinks that she is clever. (Freidin 1997)

One possibility is to provide an addendum to the interpretive procedure, much like the

one stated for (a). That is, one could revise the rule as:

(b) If α is a pronoun, interpret it as disjoint from every c-commanding phrase in D. If

there is a non-c-commanding NP or a c-commanding NP outside of D, the pronoun may

be interpreted as co-referential with that NP only if the NP is an eligible antecedent.

While this rule accounts for possible coreference of a pronoun and another NP within the

sentence, it still does not account for the obligatory reference of resumptive pronouns.

That is, there is no account of why resumptive pronouns must be interpreted as co-

referential with the relative head.

1.2.1.2. The Internal Structure of Pronouns

Freidin and Vergnaud (2001) propose that the underlying representation of every

definite pronoun is a definite description which includes both the gender, number, and

person features (φ-features) of the NP to which the pronoun refers as well as an

unpronounced NP element to which the pronoun refers. The general schema of the

underlying representation of a pronoun is as follows:

42

i. [DP [+def] φ NP] (Freidin and Vergnaud 2001: 653)

Both the φ-features and the silent NP of any given pronoun will be identified in the

underlying representation, but while the NP does not proceed beyond Spell-Out to PF

(but only towards LF), the φ-features are pronounced. Freidin and Vergnaud offer the

following example of how their proposal functions (Freidin and Vergnaud 2001: 654):

ii. Mary thinks [DP [+def [3rd pers., sg., fem.] Mary]] solved the problem.

iii. Mary thinks [DP [+def [3rd pers., sg., fem.] Clea]] solved the problem.

In example (ii), the pronoun is interpreted as co-referential with the subject of the

matrix clause, while in the other, the pronoun is interpreted as co-referential with some

other NP external to the matrix clause. Thus, these two sentences, identical at PF, are

distinct prior to Spell Out - one contains the NP Mary and the other contains the NP Clea.

And only when the deleted NP is identical to the higher NP in all features (including

phonetic features), can they share an interpretation. In fact, "there is an unstated

assumption that Mary and Mary have the same interpretation" (Freidin, p.c. 2012). That

is to say, in (ii), there is, in fact, a forced interpretation of the DP internal pronoun

relating it to the matrix subject.22

22

Although Freidin and Vergnaud do not explicitly state whether this model will also account for pronouns

with split antecedents, there is no reason why it cannot be extended to include those pronouns.

Mary thought that Clea would build them a new dining table.

Mary thought that Clea would build [DP +def [3rd pers., pl.] Mary, Clea] a new dining table.

The only difference between this example and the previous ones is that the pronominal DP contains a set of

referents, namely {Mary, Clea}.

43

In 2003, Cedric Boeckx advanced a theory of resumptive pronouns called the Big-

DP analysis, which closely resembles this proposal made by Freidin and Vergnaud (2001)

and which will be discussed in the following section.

1.3. Previous Analyses of Resumptive Pronouns

1.3.1. Some Early Analyses

Since the early 1980s, linguists have been trying to provide an analysis of the

phenomenon of resumptive pronouns (Doron 1982, Sells 1984, Browne 1986, Erteschik-

Shir 1992, and Shlonsky 1992 among others). Although their analyses were extremely

useful in terms of defining the nature of the relationship between a resumptive pronoun

and the relative heads (e.g. they are interpreted as the one and same), the mechanics of

their derivations suffered from the same problems as early analyses of binding: they

introduce new elements into the syntax as part of the computation.

Doron (1982) initially characterized resumptive pronouns as "overt traces". That

is, they are traces that are substituted with pronouns by means of some syntactic rule.

Some took the view that this substitution occurs as a last resort strategy to prevent the

derivation from crashing (Shlonsky 1992). In these proposals, either the trace was left by

an overt NP, namely the relative head (Doron), which raises out of its A-position, or it

was left by a silent operator, following the Chomsky (1977) model of relative clauses.

The problem with these early analyses is that they relied on formal mechanisms,

i.e. substitution of a trace by a pronoun, that are no longer viable. Assuming the

Inclusiveness Condition, the existence of traces is impossible, as they are elements with

specific syntacic properties that are generated during the course of the derivation.

44

Furthermore, these early analyses used indices to stipulate co-reference between terms,

and this was the unique mechanism whereby a resumptive pronoun would be interpreted

as co-referential with the relative head. The use of indices is also problematic given our

assumptions, as shown in section 1.2. The early analyses, while providing valuable

insight into the nature of resumptive pronouns (e.g. their obligatorily coreference with the

relative head), must be re-formulated in a way that conforms to current theoretical

assumptions.

1.3.2. Boeckx (2003) and the Stranding Analysis

The analysis of resumptive pronouns that Boeckx (2003) presents is inspired by

Sportiche's (1988) analysis of Quantifier Float, according to which a quantifier can

surface at various levels in the derivation and in various positions with regard to the

quantified N (e.g. All the students left versus The students all left). Sportiche's theory of

how quantifier float arises is that the quantifier enters the derivation as a Q sister of the

NP it quantifies; that is, all the students would be inserted into the syntax as a single

constituent QP. The entire QP may then raise to higher phrasal positions to satisfy EPP

requirements. Alternatively, the NP may raise out of the QP, thereby stranding the

quantifier either in its original position or in any of the intermediate positions which it

occupies. Boeckx suggests that resumption is derivationally analogous to Quantifier

Float; the resumptive pronoun and the NP to which it refers are first merged to form a

constituent. The antecedent NP then raises first to Spec-DP and subsequently to Spec of

the relative CP, leaving the resumptive pronoun behind.

45

This proposal is similar to many earlier proposals, notably the analysis of clitic-

doubling23

first introduced by Kayne (1972) and subsequently adopted and reanalyzed by

many scholars according to changing models of syntax (Torrego 1986, Uriagereka 1988,

Cecchetto 2000, among others). The basic concept, however, remained the same: the

clitic and the NP it refers to are initially in a sisterhood relationship in the derivation.

Boeckx maintains Cechetto's (2000) analysis of the derivation of clitic doubling and

extends it to handle resumption as well. In Cechetto's structural analysis, repeated below,

the clitic and its sister NP are co-referential:

Diagram 8

DP

D'

clitic NP

This is identical to the structure proposed by Boeckx, except that instead of a clitic

double, the resumptive pronoun is the determiner head. Then, in a process analogous to

Sportiche's quantifier float, the NP is extracted from the DP whose head, the resumptive

pronoun, is stranded. Thus the DP structure containing the resumptive pronoun ends up

23

Clitic doubling occurs in many languages (e.g. Spanish, Macedonian, Albanian) in which a R-expression

is "doubled" by a clitic pronominal form, as in the Macedonian example below:

Macedonian

Ja pročitav kniga-ta za volšebnik-ot.

herFEM.SG.ACC I-read1ST.SG.AOR bookFEM.SG.-theDEF. about wizard-the

'I read (it) the book about the wizard.'

The direct object of V pročitav, meaning read, is knigata, meaning the book. In Macedonian, a definite

direct or indirect object NP must additionally accompanied in the syntax by a clitic pronoun that matches it

in all syntactic features. Thus, the clitic pronoun ja appears.

Steven Franks and Catherine Rudin (2005) offer an account for this phenomenon in Bulgarian.

They propose that the DP is nested within another phrasal shell, namely a K(ase)P. The clitic pronoun is the

head of KP and takes DP as its complement. The DP embedded within the KP is the associate of the clitic

pronoun. While I restrict myself to a brief descriptive footnote here, I will return to the topic of clitic

doubling in chapter 3.

46

looking very much like the DP structure of all definite pronouns proposed by Freidin and

Vergnaud (2001):

(29) [DP [[resumptive pronoun] φ NP]]

Unlike the Freidin and Vergnaud proposal, however, the NP is not elided, but rather is

raised out of the DP into another position higher in the clause. Boeckx's analysis of

resumptives is tied to the Vergnaud-Kayne model of relative clauses; after the relative

head NP has merged with the resumptive pronoun D, the resulting DP is merged into an

A-position within the relative clause. The relative head NP is then extracted from the DP

and raises first to Spec-DP and subsequently to Spec-CP of the relative clause.

Diagram 924

CP

NP ...

DP

NP DP

D NP

resumptive pronoun relative head

Boeckx's proposal was highly influential, and since its publication, there has been

wide acceptance that the resumptive pronoun is part of a larger phrase in which it is first

merged with its co-referential NP. In nearly every subsequent generative analysis of

resumptive pronouns, it has been assumed that they are initially merged with their

24

Note, that the structure shown here is highly simplified. All of the intermediate nodes between the matrix

relative clause CP and the DP containing the resumptive pronoun are omitted.

47

antecedents to form a structure that is then merged into the relative clause. Nevertheless,

Boeckx's analysis does present some problems: 1) It does not offer a unified account of

pronouns with resumptive pronouns; he treats resumptive pronouns as lexically distinct

from ordinary pronouns. 2) Because it relies on the Vergnaud-Kayne analysis of relative

clauses, it results in a case conflict for the relative head. 3) His analysis fails to provide

an account of the optionality of resumptive pronouns25

.

While I adopt many elements of Boeckx's analysis (most significantly, his

analysis of resumption as the stranding of a pronoun by its complement), I present an

alternative to the structure that he proposes. The new analysis which I develop in the next

chapter is rooted in three sources: 1) Boeckx's resumption as stranding theory, 2) the

Sauerland matching model of relative clauses. 3) the pronominal structure proposed by

Franks and Rudin (2005) in which the pronoun takes a DP complement headed by a

definite determiner (see footnote 23).

25

In some languages (e.g., Hebrew, B/C/S) resumptive pronouns are optional (i.e. they may or may not be

pronounced) under certain conditions. This will be discussed further in Chapters 3 and 5.

48

CHAPTER 2

A NEW ANALYSIS OF CANONICAL RESUMPTIVE PRONOUNS

In this chapter I offer a new analysis of canonical resumptive pronouns, which

incorporates many of the elements of previous analyses, namely a version of Sauerland's

(2003) matching model of relative clauses, a modification of Boeckx's (2003) stranding

analysis of resumptive pronouns, and the Franks and Rudin (2005) analysis of clitic

doubling. However, my general approach differs from previous approaches in two crucial

ways: First, I treat resumptive pronouns as no different from ordinary pronouns. The

consequence of this assumption is twofold: I must develop a uniform structural analysis

of pronouns as well as a unified theory of pronominal interpretation, both of which must

extend to all instances of pronouns26

. Second, I develop an analysis that not only formally

describes the conditions under which resumptive pronouns may occur, but one that can

predict their occurrence. In the following chapter my goal is to identify the source of

resumption as well as some of the universal and parametric constraints on it.

2.1. Requirements on the Derivation

The syntactic analysis of canonical resumption presented in this chapter is

structured to satisfy four specific derivational requirements. The first of these has to do

with the status of clitic pronouns with respect to long-form pronouns. I argue that the two

pronominal forms are, in fact, not distinct in their interpretation but only in terms of a

feature that does not allow clitics to occur under phonological stress. Thus, the first

requirement is that clitic pronouns must not occur in focus positions.

26

This includes all three varieties of pronouns (clitic, deficient, and strong) identified in Cardinaletti and

Starke (1999).

49

Furthermore, because I assume that resumptive pronouns and ordinary pronouns

are simply different instantions of the same lexical item, they enter the derivation with the

same set of syntacic features to check. One of these features is definiteness27

. The second

structural requirement, which will be discussed in 2.1.2, is that pronouns must check their

[def] feature with another syntactic item (a definite N, a definite NP, or a definite DP).

This means that definiteness is not a lexical primitive of pronouns, but like their φ-

features, it is something obligatorily derived through checking relationships with other

syntactic elements.

In section 2.1.2.1, I apply a diagnostic from Macedonian to show that the moved

relative head within a relative clause is invariably definite. Although the relative head at

PF may be indefinite, the moved phrase of the relative clause is always interpreted as

definite. This leads to the conclusion that there are two distinct heads (as proposed by

Sauerland 1998, 2003). The third structural requirement, then, is that there must be two

distinct heads and that the internal head must be definite.

Finally, the fourth requirement, which will be discussed in Section 2.1.3, is that all

personal pronouns, including those that are resumptive (i.e. stranded by movement), must

be subject to the same lexical and interpretive rules. That is to say, while we distinguish

resumptive pronouns from ordinary pronouns based on the syntactic configurations in

which they appear, they are actually instances of the same lexical item, and thus the same

binding criteria apply to them.

27

I am intentionally leaving out the "generic" pronoun that can occasionally be found in English proverbs:

Let him who is without sin cast the first stone. (John 8:7)

This pronoun is actually different from both resumptive and ordinary personal pronouns. It essentially has

the same meaning as anyone; it has no specific referent.

50

2.1.1. The Status of Clitic Pronouns

One robust generalization we find is that, in languages that have clitic pronominal

forms, those are the forms that occur as resumptives28

. While I do not make a distinction

between resumptive pronouns and ordinary pronouns, clitic pronouns do differ from

long-form pronouns in their syntactic features. Clitics have two features that set them

apart from long-form pronouns: First, they cannot ocur under stress, and therefore they

can never move into any focus position. Second, they have certain prosodic features that

require them to be pronounced in a specific position in the linear sequence of the

sentence. In Slavic, this is often the second position (also known as the Wackernagel

position)29

. I will assume that the clitic position is determined by a PF linearization rule

(as in Richards 2002, Franks 2008).

Nevertheless, the interpretation of clitic pronouns is not distinctive, i.e. there is no

evidence supporting any LF-legible feature besides focus that distinguishes clitic

pronouns from their long-form counterparts. In most cases, they are interchangable30

:

28

It is not, however, the case that resumptive pronouns uniquely surface as clitics. In Hebrew, which has

only one set of pronouns (that is, it does not have a distinct set of clitic pronominal forms), the ordinary

form of the pronoun occurs:

Hebrew

Ha-ʔiš še- raʔiti ʔoto.

the-man that I-saw him

'the man that I saw (him)' (Shlonsky 1992)

I will address this data in section 2.2. 29

What qualifies as second position in the linear sequence and the operation by which a clitic moves into

that position are much debated topics among Slavists. As it is both a controversial question and one that is

not immediately germane to my analysis, the details of clitic movement will not be included as part of this

thesis. 30

One general difference between clitic pronouns and long-form pronouns is that clitic pronouns must be

bound; they can never occupy positions where a null pronoun could potentially occur. Long-form pronouns,

on the other hand, are not necessarily bound. Montalbetti (1984) gives evidence that an overt pronoun

cannot be bound if it occurs in a position that could be occupied by a null pronoun.

51

Slovene

(30) Ana pozna njega.

AnaNOM.SG.FEM. know3rd.SG.PRES. himACC.SG.MASC.

'Ana knows him.'

(31) Ana ga pozna.

AnaNOM.SG.FEM. himACC.SG.MASC. know3rd.SG.PRES.

'Ana knows him.'

Because clitic pronouns have a PF-feature that prohibits them from being stressed, they

cannot occur in focus positions. On the other hand, long-form pronouns are used

exclusively in focus positions31

. That is, (30) and (31) differ from one another because

the use of the long-form pronoun in (30) implies contrastive stress; Ana knows him and

not someone else. Such contrastive focus is ruled out in (31).

One can then posit that the only feature setting clitics apart from long-form

pronouns is that the latter have a [+foc] feature that causes them to move into focus

positions. Clitic pronominal forms, on the other hand, have a [-foc] feature that prevents

them from moving into focus positions. The second structural requirement amounts to a

restriction on clitic pronouns prohibiting them from moving into focus positions.

31

Prepositions also tend to take only the long-form of a pronoun as an object. The reason for this is that

prepositions, like clitic pronominals, frequently do not bear stress. They form a single accentual unit with

their object which is why the object must be an element that can bear stress. While a clitic pronominal

cannot bear stress, a long-form pronominal can. That said, there are instances of the preposition bearing

stress and forming an accentual unit with a clitic object pronoun, as in Slovene (data taken from a Slovene

online magazine, www.zurnal24.si):

Mussomeli poziva Američane, naj glasujejo za-nj.

MussomeliNOM.SG.MASC. calls AmericansACC.PL. in-order-that they-vote3rd.PL.PRES. for-himACC.SG.MASC.

'Mussomeli calls upon Americans to vote for him.'

The term zanj is the combination of the preposition za, meaning for, with a clitic form (-nj) of the pronoun

njega, meaning him. Here, the stress is on the preposition, so it is able to host a clitic pronoun.

52

2.1.2. Checking Definiteness on the Pronoun

Freidin and Vergnaud (2001) postulate a [+def] feature which, when combined

with a complete set of φ-features, spells out as a pronoun. Extending their proposal into a

syntactic operation, I suggest that a pronoun must check its definiteness feature with

another syntactic item. The second requirement on the structure is that personal pronouns

(both clitic and long-form) must check their definiteness feature during the course of the

derivation. They can either check this feature either with a definite NP (i.e. a proper

name or a geographical name) or with definite Determiner (a definite article or a D-linked

wh-word).

Diagram (10) illustrates a pronoun whose definiteness feature can be checked

with a definite NP:

Diagram 1032

DP

D NP

pro[+def] Mary[+def]

Diagram (11) shows the pronoun can check its definiteness feature against a definite DP:

Diagram 11

DP

D DP[+def]

pro[+def]

D NP

the[+def] book[+def]

32

One can argue that proper nouns are also DPs in which the head N moves into D, as claimed by

Longobardi (1994), but this claim does not present any critical consequences on the theory proposed here.

53

Thus, the pronoun checks its definiteness through a sisterhood relationship with either NP

or DP. Similary, it checks its φ-features with those of either its NP sister or the NP

immediately dominated by its DP sister.

Under most conditions, the NP or DP sister of the pronoun is deleted. While this

process, namely the deletion of the NP, is assumed to occur in Freidin and Vergnaud's

proposal, the formal mechanics of the operation are not specified; rather they simply refer

to the NP as a "silent NP component" (Freidin and Vergnaud 2001: 653) of the pronoun.

I propose the following PF condition:

(32) Pronominal Associate Deletion: if D0 is a personal pronoun and XP is a sister to D

0,

then XP must be deleted.

Because this deletion is a PF operation, the NP/DP must still be a sister to the pronoun at

PF. That is, if the associate raises to a position higher than the pronoun prior to Spell-Out,

thereby stranding the pronoun (as per Boeckx 2003), this deletion operation cannot occur,

and thus both the pronoun and the NP/DP with which it was first merged will be

pronounced.

This is, in essence, the source of resumptive pronouns - they are simply ordinary

pronouns whose internal NPs/DPs have raised into a higher projection.

54

2.1.2.1. The Definiteness of Relative Clause internal DPs: Evidence from Macedonian

Clitic Doubling33

In the Vergnaud-Kayne analysis of relative clauses, the relative head first merges

into the derivation as a bare NP without any determiner. The determiner that is eventually

linearly adjacent to the relative head is actually a D0 sister to the entire relative clause CP.

This analysis of the syntax does not accurately reflect the fact that the gap within a

relative clause is uniformly interpreted as definite.

The invariant definiteness of the relative clause gap can be shown by applying a

diagnostic using data from Macedonian, which exhibits clitic doubling only for definite

objects.

Macedonian

(33) Pročitav edna kniga od Petre Andreevski včera.

I-read1st.SG.AOR. aACC.SG.FEM. bookACC.SG.FEM. by Petre Andreevski yesterday.

'I read a book by Petre Andreevski yesterday.'

Edna in (33) is an indefinite adjective34

, which takes scope over kniga. The NP edna

kniga is the direct object of V pročitav.

Unlike most other Slavic languages (but much like English), Macedonian R-

expressions do not inflect for case, but they are marked for definiteness by an enclitic

article. When the direct object is definite, the sentence must include an additional clitic

pronominal element - a clitic double of the direct object.

33

In this section I do not provide an analysis of clitic doubling, only a description of the phenomenon. In the

next chapter, I will address clitic doubling in detail, as it, too, is a case of resumption. 34

NPs are automatically interpreted as indefinite in Macedonian, unless an NP is embedded in a DP headed

by a definite D0. I am assuming that edna is not a determiner but rather an indefinite adjective (see

Dimitrova-Vulchanova and Giusti 1998: 357).

55

(34) *(Ja) pročitav kniga-ta od Petre Andreevski včera.

itACC.SG.FEM. I-read1st.SG.AOR. book-the by Petre Andreevski yesterday.

'I read (it) the book by Petre Andreevski yesterday.'

The clitic ja matches the direct object knigata in φ-features but additionally exhibits case

morphology; Macedonian, like English, shows morphological case only on pronouns.

The clitic pronoun in (34) is obligatory, whereas in (33), a clitic double would be

ungrammatical. The crucial difference is that in (33), the direct object is indefinite while

in (34) it is definite. Clitic doubling only occurs in Macedonian when the object is

definite. Therefore, we can use Macedonian as a test for the definiteness of an object by

determining whether or not a clitic double is required35

.

(35) Na policata beše edna kniga, [RC što *(ja) napiša

on the-shelf was aSG.FEM. bookSG.FEM. that itACC.SG.FEM. wrote3rd.SG.AOR.

Petre Andreeveski].

Petre Andreevski

'On the shelf there was a book that Petre Andreevski wrote (it).'

Sentence (35) contains a relative clause, whose relative head, edna kniga, is indefinite.

However, there is an obligatory clitic within the relative clause that matches the relative

head in φ-features, just as the clitic double in (34) matches the definite direct object in φ-

features. Unlike in (34), however, the clitic differs from its overt antecedent, kniga, in its

case. While kniga is nominative, the pronoun ja is accusative. Moreover, kniga is

indefinite, and in Macedonian, indefinite objects cannot have clitic doubles. Thus, the

pronoun ja must be the clitic double of the unpronounced direct object of napiša. If this

unpronounced object were simply a copy of edna kniga, it would be indefinite, in which

35

Many thanks to Steven Franks, who suggested this test for definiteness.

56

case, the unpronounced object of the V napiša could not have a clitic double. Because the

clitic double is present in the relative clause, we can gather that its "associate" (Franks

and Rudin 2005) must be definite. Therefore the object of the relative clause internal

predicate must be distinct from the pronounced relative head, indicating that the

Sauerland model of relative clauses is likely the correct one.

Heim (1987) provides independent evidence that the pronoun in (35) is definite.

She shows that overt bound variable pronouns are always definite by running Milsark's

(1977) Definiteness Restriction diagnostic36

. She presents the following data as evidence:

(36) * Few people admitted that there had been them at the party.

(37) * No perfect relationship is such that there is it.

(Heim 1987)

Although their antecedents (i.e. few people and no perfect relationship) are indefinite, the

overt bound variable pronouns are prohibited from following there-be. Similarly, in (35),

while the apparent antecedent of ja is the indefinite NP edna kniga, the pronoun itself

must be definite.

These data not only suggest that Sauerland's model in which there are two relative

heads (one internal and one external) is correct, but furthermore, they reveal that the

internal relative head must be definite regardless of the definiteness of the external head.

36

The Definiteness Restriction prohibits definite NPs from occurring as the logical subject of existential

sentences:

* There is the horse in the barn.

Definites will only work in these sentences when they are part of a pair-list series:

There is the horse in the barn and the chickens in the coop and the dogs in the yard.

57

2.1.3. The Interpretation of Resumptive Pronouns

The interpretive procedure for pronouns in Section 1.2.1.1 and restated below is a

slightly revised version of the rule initially proposed by Chomsky and Lasnik (1993) and

subsequently elaborated by Freidin (1997):

(b) If α is a pronoun, interpret it as disjoint from every c-commanding phrase in D. If

there is a non-c-commanding NP or a c-commanding NP outside of D, the pronoun may

be interpreted as co-referential with that NP only if the NP matches the pronoun in φ-

features.

This rule (which is a re-formulation of Condition B), given a domain of CP, explains the

interpretive relationships of pronouns in non-resumptive positions but falls short of

explaining the obligatory interpretation of canonical resumptive pronouns as coreferential

with the relative head; the interpretive rule (b) allows any pronoun to have arbitrary

reference at the sentential level.

(38) Harryi thought Snapej saw himi//*j/k in the corridor.

In (38), the pronoun him may refer to Harry, but crucially cannot refer to Snape: Snape is

within the CP domain of the pronoun and c-commands it, thus their co-reference would

constitute a violation of the interpretive procedure (b). But the interpretation of the

pronoun is nonetheless ambiguous; him could refer either to Harry or to some other

individual not mentioned in the matrix clause (e.g. Ron), deriving its interpretation

through discourse. A simple proof of this lies in the grammaticality of sentence (39).

58

(39) Harryi thought Snapej saw her*i/*j/k in the corridor.

Because there is a gender feature mismatch between the pronoun and its only available

antecedent (i.e. Harry), the reference of the pronoun cannot be determined in the sentence

but rather must depend on some element outside the matrix clause (i.e. in the discourse).

Although the pronoun's reference cannot be determined sentence internally, (39) is

nevertheless grammatical. This is precisely what we expect, given interpretive rule (b).

In sharp contrast to the free interpretation of ordinary pronouns, canonical

resumptive pronouns are obligatorily bound by the relative head and thus cannot refer to

any other potential antecedent.

Slovene

(40) Študenti je videl profesorjaj, ki so studentNOM.SG.MASC. AUX3rd.SG. sawPAST.SG.MASC.. professorACC.SG.MASC. that AUX3rd.PL

mu*i/j/*k dali nagrado.

himDAT.SG.MASC. gavePAST.PL. awardACC.SG.FEM.

'The student saw the professor that they gave the award to (him).'

It is impossible for the dative pronoun mu in the relative clause in (40) to refer to any

individual other than the professor. The pronoun cannot refer to the student, even though

they share the same φ-features and therefore could potentially be co-referential under

other circumstances.37

The pronoun is also unable to refer to some other individual

37

Študenti je videl profesorjaj, ki mui/j/k je

studentNOM.SG.MASC. AUX3rd.SG. sawPAST.SG.MASC. professorACC.SG.MASC. that himDAT.SG.MASC. AUX3rd.SG.

dal nagrado.

gavePAST.SG.MASC. awardACC.SG.FEM.

'The student saw the professor that he(student) gave the award to.'

'The student saw the professor that he(discourse) gave the award to.'

'The student saw the professor who gave him(student) the award.'

'The student saw the professor who gave him(discourse) the award.'

59

mentioned in the discourse, as illustrated by the absence of any interpretation of the

pronoun in (41).

(41) * Študenti je videl profesorjaj, ki so

studentNOM.SG.MASC. AUX3rd.SG. sawPAST.SG.MASC. professorACC.SG.MASC.that AUX3rd.PL.

ji*i/*j/*k dali nagrado.

herDAT.SG.FEM gavePAST.PL. awardACC.SG.FEM.

'The student(male) saw the professor(male) that they gave the award to (her).'

Sentence (41) stands in stark contrast to (39), in which none of the NPs within the matrix

clause is an eligible antecedent of the pronoun but which is nevertheless grammatical

because the pronoun can refer to some individual mentioned previously in the discourse.

There is no possible interpretation of (41): the absence of a potential antecedent in the

clause renders the sentence ungrammatical. However, even the presence of another

potential antecedent NP (i.e., possessing matching φ-features) in the clause will not repair

the sentence, if that potential antecedent is not in the position of the relative head.

In the example above, it is possible for the dative pronoun, mu, to take študent as its antecedent. In

fact, this sentence is ambiguous between four potential readings; one in which the student gave the

professor a reward, one in which some other person X (presumably identified in the discourse) have the

professor the award, one in which the student is the recipent of the award given by the professor, the other

in which some other person X (presumably identified in the disccourse) is the recipient of the award given

by the professor.

The source of this ambiguity is that the dative pronoun mu could either be a resumptive pronoun

(that is, a pronoun that has been stranded) or an ordinary pronoun. Slovene is a subject pro-drop language

and resumptive pronouns never surface in subject position. Thus, it is unclear whether it is the dative object

being relativized and the subject is simply null due to the subject-pro drop nature of Slovene or whether it

is the subject that is being relativized and the dative object is simly an ordinary (i.e. non-resumptive)

pronoun.

60

(42) * Študentkai je videla profesorjaj, ki so

studentNOM.SG.FEM. AUX3rd.SG. sawPAST.SG.FEM. professorACC.SG.MASC. that AUX3rd.PL.

herDAT.SG.FEM. gavePAST.PL. awardACC.SG.FEM.

ji*i/*j/*k dali nagrado.

'The student(female) saw the professor(male) that they gave the award to (her).'

If a feature mismatch prevents co-reference between the resumptive pronoun and the

relative head, the resulting sentence is ungrammatical because a canonical resumptive

pronoun invariably refers to the relative head.

Furthermore, a resumptive pronoun cannot be bound by a split antecedent, even if

the relative head is part of that antecedent. Consider the contrast between (43) and (44)

below.

Slovene

(43) Irinai je povedala Jožetuj, da njenii starši

IrinaNOM.SG.FEM. AUX3rd.SG. toldSG.FEM.PAST. JožeDAT.SG.MASC. that her parentsNOM.PL.

so jimai+j prepovedali živeti skupaj.

AUX3rd.PL. them3rd.DAT.DU. forbidPL.PAST. to-live together

'Irinai told Jožej that heri parents forbid themi+j to live together.'

The pronoun in the subordinate clause in (43), jima, can take both the subject and the

dative object of the matrix clause together as its antecedent: the antecedent of the jima is

the non-contiguous pair {Irina, Jože}. This is evident from the form of the pronoun,

which indicates a Dual number.

61

(44) * Irinai ima-rada človekaj, ki so jimai+j njenii

IrinaNOM.SG.FEM. loves manACC.SG.MASC. that AUX3rd.PL. them3rd.DAT.DU. her

starši prepovedali živeti skupaj.

parentsNOM.PL. forbidPL.PAST. to-live together

'Irinai loves the manj that heri parents forbid themi+j to live together.'

In (44) the relative clause internal canonical resumptive pronoun cannot be anaphoric on

a non-contiguous antecedent; it can take only the relative head as its antecedent. A natural

question to ask given these data is how it will possible to account for the restriction

prohibiting split antecedents of resumptive pronouns while allowing split antecedents

generally. This question will be addressed in Section 2.2.4. But we must first address a

related question, namely, how to account for the obligatory reference between the

resumptive pronoun and the relative head.

If we adopt Boeckx's (2003) proposal that a pronoun's referent is an NP internal to

the DP headed by the pronoun, we can propose that a pronoun must be interpreted as co-

referential with the

NP internal to the DP headed by the pronoun.

Diagram 12

DP

D NP

proi[φ] Ri[φ]

In this structure, pro represents any personal pronoun and R is an R-expression. Whereas

Boeckx posits this structure exclusively for resumptive clitic pronouns, I follow the idea

proposed by Freidin and Vergnaud (2001) that the referent NP is part of the underlying

62

structure of every definite pronoun. Freidin and Vergnaud (2001:654) furthermore

propose that "the unpronounced NP component of the pronoun determines its

interpretation". Because I have extended their proposal into a syntactic procedure, I must

provide a formal condition accounting for the interpretation of the pronoun.

I propose the following LF condition:

(45) Pronominal Reference Condition: In DPMAX

whose D0 is a pronoun, interpret D

0 as

co-referential with the NP also dominated by DPMAX

.

This will also account for the interpretation of pronouns that have a DP sister38

, as in

diagram (11) in section 2.1.2, repeated below:

Diagram 11

DP

D DP[+def]

proi[φ,+def]

D NP

D[+def] Ri[φ]

As the derivation proceeds, there may be another element merged into the derivation

possessing the same interpretation as the pronoun and its coreferential NP, thus giving the

coreferential interpretation of sentence (28), repeated here:

(28) Maryi thinks that she(Mary)i is clever.

38

Because the Pronominal Reference Condition applies at LF, if the internal DP of the structure shown here

raises to a higher position prior to Spell-Out, it is the copy of NP that gives the pronoun its reference.

63

Alternatively, there may be no other element in the derivation possessing the same

interpretation as the pronoun, as in example (39). So long as any phrase that does share

the same interpretation as the pronoun does not c-command the pronoun within the CP

domain, no aspect of the interpretive rule (b) is violated.

Thus one of the constraints on the derivation in which canonical resumption

occurs is that the relative head, with which the resumptive pronoun is interpreted as

coreferential, must be in a position where it either does not c-command the pronoun or

where it is outside of the relative CP. This falls out naturally from Sauerland's matching

model of relative clauses; the external head is merged outside of the relative CP, allowing

it to be interpreted as co-referential with the relative-CP-internal pronoun. What forces

this coreference, however, is the following general condition on relative clauses: the

external head must be the antecedent of the internal head (Sauerland 2003).

2.2. A New Analysis

2.2.1 The Structure of the Relative Clause and the Structure of the Pronoun

2.2.1.1. The Matching Model

As shown in chapter 1, the matching model of relative clauses proposed by

Sauerland (1998, 2003) is essentially a hybrid of the two earlier models (the Vergnaud-

Kayne raising analysis and the Chomsky adjunction analysis). But since it employs facets

of both, it has many advantages over each of those individual analyses. It was also shown

above that an analysis in which there are two heads, one external and one internal to the

relative clause, accounts for a larger set of data.

By employing Sauerland's matching model, we need not stipulate ancillary

interpretive rules applying uniquely to resumptive pronouns. They can be interpreted

64

according to condition (b) just as an ordinary pronoun, as will be shown in section 2.2.2.

In addition, Sauerland's model provides a means to account for a possible difference in

definiteness between the internal relative head and the external relative head.

Therefore, for the new analysis proposed here, I will adopt Sauerland's matching

model, in which there are two heads. I make only one adjustment to his theory: whereas

Sauerland (2003) simply states that the external relative head must be the antecedent of

the internal relative head and that the internal relative head must be deleted, I provide a

more specific rule by which this deletion applies.

(46) Relative Deletion (slightly altered): In relative clauses, the internal head must take

the external head as its antecedent. Any lexical item in Spec-CP of the relative clause that

takes the external head as an antecedent must be deleted.

2.2.1.2. The Stacked-DP Pronoun Structure

In section 2.1.2, I presented two configurations in which a pronoun may occur: in

one, there is a D pronoun whose sister is an NP and in the other the D pronoun is the head

of one DP which immediately dominates another definite DP. It is only the second of

these configurations, which I call a stacked-DP, that can give rise to resumptive

pronouns39

. Resumption, which is simply another way of saying the stranding of a

pronoun by its internal DP, can only occur in this (i.e. stacked-DP) configuration,

because an NP cannot raise out of the minimal DP in which it is embedded. That is, an

NP cannot strand its determiner. Consider again the derivation of Quantifier Float.

39

This analysis of the resumptive pronoun is similar to the KP analysis of Macedonian clitic doubling

proposed in Franks and Rudin (2005). The primary difference is that in my analysis, there is no head K(ase)

designated specifically for clitic pronouns; all pronouns are D and clitics are unexceptional.

65

Diagram 12

QP

DPi QP

D NP Q DPi

the students all

D NP

the students

The DP the students moves into the Spec-QP, giving the surface word order the students

all. It is impossible, however, for the NP to move into the Spec-QP stranding the

determiner, which would give *students all the. The NP must pied-pipe the determiner

with it into the higher phrase. It is, however, possible to extract a DP from within another

DP, in much the same way as the DP is extracted from QP in Quantifier Float.

In both ancient and modern Greek there are instances of DP-stacking40

. The

following is a famous example from ancient Greek, in which the N0 occurs between a

definite article and a demonstrative, both of which agree with the N0 in accusative case.

Ancient Greek41

(47) ouk oida ton anthrōpon touton

not I-know theACC.SG.MASC. manACC.SG.MASC. thisACC.SG.MASC.

'I do not know this man. '

(Apostle Peter's denial of Christ, Ioannes Damscenus, fragmenta e cod. Vat. gr. 1236)

40

This is following Panagiotidis (2000), who proposes that demonstratives are, in fact, D0 heads and that

they may have some further internal structure (i.e. they may constitute a simplex DP), arguing against

earlier theories (e.g. Stavrou and Horrocks (1989), among others) that the demonstrative occupies Spec-DP.

But while Panagiotidis allows for the possibility that the demonstrative could be a full phrase, I propose

that it is a head. 41

Many thanks to Jozef Müller for the ancient Greek data.

66

Below are examples of DP-stacking in Modern Greek.

Modern Greek42

(48) to vivlio (50) *auto vivlio

the book this book

'the book' ‘this book’

(49) auto to vivlio (51) *vivlio auto

this the book book this

'this book' ‘this book’

(48) shows that not all Greek DPs are stacked DPs. However, as illustrated by

ungrammaticality of (50) and (51), Greek demonstrative determiners obligatorily cooccur

with a definite determiner. Thus, Greek requires that determiner D heads be merged with

another D. This occurs for the same reason that a pronoun D must be merged with

another D: to check a feature. While Panagiotidis (2000) suggests this feature is [ref], I

suggest that it is the definiteness feature, [def], that must be checked. Diagram (13)

shows the structure of sentence (49):

Diagram 13

DP

D DP

this

D NP

the book

Once the demonstrative checks its features with the internal D head, the entire internal

DP can then raise into the Spec position of the higher DP headed by D this. This

42

Many thanks to Phoevos Panagiotidis and Pavlos Avlamis for the modern Greek data and judgements.

67

movement operation is analogous to what we find in Quantifier Float and results in the

structure in diagram (14).

Diagram 14

DP

DPi DP

D NP D DPi

the book this

D NP

the book

The definite article followed by the NP followed by the demonstrative is an attested

sequence in modern Greek (see (52)) as well as ancient Greek (see (47)).

(52) to vivlio auto

the book this

'this book'

Thus, the structure I suggest for pronouns is not unique; DP-stacking occurs in several

languages (e.g. Greek, Macedonian, Hungarian) under other conditions as well.

The movement of the embedded DP is crucial in the analysis of resumptive

pronouns, because it strands the pronoun, thereby forcing both the pronoun as well as the

moved DP to be pronounced (i.e., eliminating the conditions under which Pronominal

Associate Deletion may apply).

68

2.2.1.3. Languages with and without Canonical Resumption

The analysis of resumptive pronouns I have outlined above reduces the

differences between languages with and without resumption to a single parametric

difference: if the internal DP of a stacked-DP can move independently, the language will

have resumptive pronouns. If the internal DP must pied-pipe the pronoun, the language

will not have resumptive pronouns. Given this generalization, there is no reason to posit

additional differences between relative clauses in languages with and without resumptive

pronouns. I suggest that cross-linguistically the relativized object is a pronoun whose

sister contains the internal head of the relative clause. Thus, some languages will exhibit

canonical resumption and others will not, depending only upon the parametric difference

of internal-DP movement versus pied-piping43

.

2.2.1.4. Island Effects

This movement analysis of resumptive pronouns inevitably raises questions

pertaining to the consequences of this movement from within an island. One would

predict that the presence of a resumptive pronoun at the origin site of a movement

operation would have no effect on any island violation resulting from that movement. In

other words, if a movement operation constitutes an island violation (thereby producing

an ungrammatical sentence), then the presence of a resumptive pronoun stranded by that

movement should not ameliorate those island effects.

43

It is not necessarily the case that all languages whose internal DP is independently mobile (i.e. languages

in which resumption is possible) will show the same overt stacking of determiners that occurs in Greek. In

many languages, the combination of a demonstrative and an article is ruled out for reasons independent of

structural constraints (e.g. rules relating to parsimony: the demonstrative automatically conveys

definiteness, so an additional definite determiner is redundant).

Likewise, a language that has overt stacked-DPs will not necessarily have resumptive pronouns, if

the internal DP in that language is not independently mobile (this will be shown explicitly in section

2.3.1.1.)

69

In some languages, such as B/C/S, this prediction proves correct: an island

violation (weak or strong) cannot be remedied by the presence of a resumptive pronoun.

(53) * Čovek, koga se sećam gde

manNOM.SG.MASC. whomACC.SG.MASC. REFL remember1st.SG.PRES. where

sam upoznala.

AUX1st.SG.PRES. metSG.PAST.FEM.

'the man whom I remember where I met him'

(54) * Čovek što se sećam gde sam

manNOM.SG.MASC. that REFL remember1st.SG.PRES. where AUX1st.SG.PRES.

ga upoznala.

himACC.SG.MASC. metSG.PAST.FEM.

'the man that I remember where I met him' (Boeckx 2003)

Both of these sentences contain a wh-island from which, according to the analysis

I have presented, the internal relative head must raise. In sentence (53), the DP containing

the D0 koga and the internal relative head raises past the wh-phrase gde to Spec-CP of the

relative clause. This movement constitutes a wh-island violation, and the sentence is thus

ungrammatical. An identical movement operation occurs in sentence (54), only the

moved elements differ: the DP containing a null D0 operator and the internal relative head

raises to Spec-CP. As predicted, this movement, too, constitutes a wh-island violation.

In many other languages, however, the prediction does not hold. While movement

out of an island by a DP headed by an overt D0 results in an island violation, the same

movement made by a DP headed by a PF-null D0 (and the subsequent stranding of a

resumptive pronoun) will be grammatical. Consider the following Slovak data:

70

(55) * Človek, ktorého si spomínam kde

manNOM.SG.MASC. whomACC.SG.MASC. REFL remember1st.SG.PRES. where

som stretol.

AUX1ST.SG.PRES. metSG.PAST.MASC.

'the man which I now remember where I met'

(56) Človek, čo si spomínam kde som

manNOM.SG.MASC. that REFL remember1st.SG.PRES. where AUX1ST.SG.PRES.

ho stretol

himACC.SG.MASC. metSG.PAST.MASC.

'the man that I remember where I met him'

Boeckx (2003) examines this question in great detail and presents a varied array

of relevant empirical data. While in some languages, all movement out of islands results

in ungrammatical sentences, in others island violations appear to be ameliorated by the

presence of a resumptive pronoun. Furthermore, Boeckx shows that some languages can

rescue certain island violations (weak islands) by the introduction of a resumptive

pronoun while strong island violations will result in an ungrammatical sentence

regardless of the presence of a resumptive pronoun.

Boeckx's solution to this complex problem is to tease apart two syntactic

operations, namely Match and Agree. His proposal, put very simply, is that resumptive

chains are the result of a Match operation, and chains created by means of Match are

impervious to island conditions. Chains built on the Agree operation, on the other hand,

are island-sensitive. Boeckx is quick to point out that in some languages (e.g. B/C/S),

resumptive chains involve Agree, and therefore those chains will be island-sensitive. In

many ways, his proposal is unsatisfactory, since the distinction he draws between

71

resumptive and non-resumptive chains is, at its core, simply a descriptive device; there is

nothing beyond the data he is attempting to explain to support the existence of such a

distinction.

The empirical data are varied with respect to resumption in islands. Languages do

not behave uniformly. While Boeckx's proposal may not be correct, he does illustrate that

"a pluralistic approach" (Boeckx 2003) is required and that any analysis of resumption

will have to contend with the broad variations in the data related to islands (i.e. a non-

movement approach to resumption would have to provide some explanation for the

island-effects in B/C/S resumptive relatives). Thus, while there is still no clear account of

how islands may be affected by the presence of a resumptive pronoun, it is evident that a

movement analysis of resumption is no weaker than a non-movement analysis based on

the island-effects alone.

2.2.2. Derivations of Canonical Resumption

In this section I present derivations based on the new analysis of resumptive

pronouns outlined above and following Sauerland's matching model of relative clauses. I

present these derivations to demonstrate how they comply with the structural

requirements noted in section 2.1 and how they account for the data.

2.2.2.1. Bosnian/Croatian/Serbian: Feature-checking and Licensing Co-reference

Diagram (15) reflects the Spell-Out syntax of sentence (1), repeated below.

72

Bosnian/Croatian/Serbian

(1) Čovjek [RC što sam ga vidio juče ]

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC. sawMASC.SG.PAST. yesterday

sjedi ovdje.

sits here

'The man that I saw (him) yesterday is sitting here.'

Diagram 15

TP

DP TP

D NP T VP

Ø

N CP V AdvP

manj sits here

DP[+def]i CP

D NP C TP

OPWH[+def] manj that

DP TP

(I)

T VP

AUX

VP AdvP

yesterday

V DP

saw

D DP[+def]i

himj[+def,-foc] D NP

OPWH[+def] manj

The NP man first merges with a silent wh-operator definite determiner. The

resulting [+def] DP is then merged with the pronoun D him, thereby creating a stacked

DP. The pronominal D0 then checks its definiteness feature with its sister DP. The stacked

73

DP then merges with V saw, which checks accusative case on its DP object. The VP then

merges with an adjunct AdvP, creating the maximal VP, which then merges with the

auxiliary. Subsequently the DP subject merges and then the complementizer, that,

merges.

Once C0 that merges, a feature on C, presumably a [wh] or [foc] feature that needs

to be checked, causes the DP sister of the pronoun (i.e. the internal DP of the stacked DP)

to raise into Spec-CP. The stacked DP itself cannot move to Spec-CP because it is headed

by a [-foc] clitic pronoun. The [-foc] feature of the head is inherited by the phrase (via

feature spreading / percolation (see Babby 1986, Franks 1995)), thus the DP headed by

the clitic is also [-foc]. The stacked DP, being marked [-foc], cannot move into focus

positions. Because Spec-CP is a focus position, the [-foc] DP headed by the clitic

pronoun is not eligible to move into it. It is only the internal DP headed by a wh-operator

that can raise into Spec-CP.

Once the DP containing the wh-operator and the internal head raise and merge

into Spec-CP, the external head N0 is merged with the relative clause CP. It is at this point

that Relative Deletion (as in Sauerland 2003) applies, i.e., the internal head, being an item

in Spec-CP (thus visible to the higher phase) that takes the external head as an

antecedent, is obligatorily deleted. This gives us the Spell-Out syntax shown in the

derivation above. At PF, the clitic moves into its dedicated position (the second, or

Wackernagel, position in many languages). This results in the correct PF word order, with

the pronominal clitic immediately following the verbal auxiliary clitic.

All the binding criteria are also satisfied. The external head N man is merged in a

position outside of the relative clause CP containing the pronoun him. Thus, the

74

interpretation of the pronoun him as co-referential with man does not violate interpretive

rule (b), which states that a pronoun may not be interpreted as co-referential with a c-

commanding phrase within D. And, following Sauerland (2003), the co-reference

between the internal head and the external head is obligatory as a condition on relative

clauses.

Finally, looking at the movement operation, one may expect to find Strong

Crossover effects, given that the a co-referential NP raises from a position lower in the

derivation to a position higher in the derivation, with respect to the pronoun. The

movement described here, however, is not movement past the position of pronoun, but

rather extraction from the pronominal DP, and therefore it does not precipitate a crossover

effect.

2.2.2.2. Hebrew: A Case of Optional Resumption

Diagram (16) represents the structure of the relative clause DP in example (57) at

Spell-Out.

Hebrew

(57) Ha-ʔiš še- raʔiti ʔoto.

the-man that I-saw him

'the man that I saw (him)' (Shlonsky 1992)

75

Diagram 16

DP

D NP

the

N CP

manj

DPi CP

D NP C TP

OPWH[+def] manj that

DP TP

I

T VP

V DP

saw

D DPi

himj

D NP

OPWH[+def] manj

With the exception of the AdvP (which is present in (1) and absent in (57)44

), the

structure of the relative clause in (57) is identical to that of the relative clause in (1). The

only significant difference between them is that the Hebrew pronoun is not a clitic form.

Unlike B/C/S, Hebrew does not have distinct [+foc] and [-foc] pronominal forms; it has

no pronominal clitics and its pronouns are unspecified for focus. This means that there is

no special feature on the pronoun prohibiting it from moving into Spec-CP. The question

this inevitably raises is: why, in Hebrew, does the movement of the internal DP

(motivated by the [+wh] feature on the operator) not also pied-pipe the pronoun into

Spec-CP? The answer is that it can. That is, in Hebrew, either the internal DP can move

44

The presence or absence of the AdvP is inconsequential to the structure of the relative clause.

76

into Spec-CP or the entire stacked DP can move into Spec-CP. Consider the following

derivation:

Diagram 17

DP

D NP

the

N CP

manj

DPi CP

D DP C TP

himj that

D NP DP TP

OPWH[+def] manj I

T VP

V DPi

saw

D DP

himj

D NP

OPWH[+def] manj

If the entire stacked DP moves into the Spec-CP, we would expect, given the revised

definition of Relative Deletion (as stated in (46)), that both the pronoun him as well as the

NP man will be deleted. The pronoun is bound by the external relative head; the two are

interpreted as co-referential. The operation of Relative Deletion specifies that any

element in Spec-CP (and therefore visible to the higher phase) that takes the external

relative head as an antecedent is obligatorily deleted. In this derivation, this operation

will force the deletion not only of the internal relative head but also of the pronoun him.

This yields the surface word order:

77

(57a) Ha-ʔiš še- raʔiti

the-man that I-saw

'the man that I saw' (Shlonsky 1992)

(57a) is, in fact, entirely grammatical: the resumptive pronoun is optional in Hebrew.

Given my analysis, this optionality is predictable. If the internal DP moves, there is a

stranded (resumptive) pronoun that is pronounced. If the stacked-DP moves, the pronoun

is obligatorily deleted, so there is no resumptive pronoun.

This is the fundamental parametric difference between languages that contain

resumptive pronouns and those that do not. If the internal DP can move independently of

the stacked DP to a higher projection, then it strands the pronoun, making it a resumptive

pronoun. If, on the other hand, the internal DP must pied-pipe the pronoun to the higher

projection, then the pronoun is ultimately deleted under Relative Deletion.

2.2.2.3. Slovene: A Case of "Clitics Only" Resumption

In this section, we explore an issue that arises in all languages that have

specialized clitic forms of pronouns. The data show that long form pronouns in languages

that also have clitic forms are never canonical resumptive pronouns. Thus, in Slovene,

(unlike Hebrew, which does not have distinct clitic forms), canonical resumptive

pronouns are always clitics. The question, then, is why the long form pronoun cannot

occur as a canonical resumptive, i.e., why is (58) grammatical while (59) is

ungrammatical?

78

(58) Človek, ki sem ga videla prihaja.

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC sawSG.FEM.PAST is-coming3rd.SG.PRES.

'The man that I saw is coming.'

(59) *Človek, ki sem njega videla prihaja.

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC sawSG.FEM.PAST is-coming3rd.SG.PRES.

'The man that I saw is coming.'

Given the DP-stacking analysis of pronouns, there is no obvious reason why the long-

form pronoun cannot be stranded. If we assume that there is only one feature

distinguishing the clitic from the long-form pronoun, namely the focus feature, then that

feature must be responsible for the difference in grammaticality between (58) and (59). In

other words, the reason that only clitic pronouns can be stranded must relate to the [-foc]

feature that is inherent in clitics.

Consider the Slovene data (30) and (31) from section 2.1.1, repeated here as (60)

and (61):

Slovene

(60) Ana pozna njega.

AnaNOM.SG.FEM. know3rd.SG.PRES. himACC.SG.MASC.

'Ana knows him.'

(61) Ana ga pozna

AnaNOM.SG.FEM. himACC.SG.MASC. know3rd.SG.PRES.

'Ana knows him.'

In (60), the long form pronoun njega is the direct object of the V pozna. In (61), the clitic

pronoun ga takes the place of the long form pronoun as the direct object. The crucial

79

difference between the two sentences has to do with focus. Sentence (60) cannot be

uttered under neutral intonation; njega is automatically prnounced under contrastive

stress. In (61) the pronoun ga, being a clitic, can never take contrastive or any other kind

of stress. The valuation of the focus feature across pronouns is then clear: while clitic

pronouns are inherently [-foc], long form pronouns are inherently [+foc].

Just as a DP headed by a [-foc] clitic pronoun is unable to raise into a focus

position, a [+foc] long form pronoun would obligatorily move into the closest available

focus position. Therefore, while the stacked DP can optionally raise into Spec of the

relative CP in Hebrew (whose pronouns are unspecified for focus), the stacked DP in

Slovene must raise into Spec-CP when the pronoun is specified [+foc]. Thus, if a clitic

pronoun is merged into Spec-CP, it will obligatorily remain in its first merge position

until Spell-Out while the internal DP raises into Spec-CP. If a long form pronoun merges,

its [+foc] feature will force it to move into Spec-CP. This gives the following structure:

80

Diagram 18

TP

DP TP

D NP T VP

is-coming

N CP

manj

DPi CP

D DP C TP

himj[+foc, +def] that

D NP DP TP

OPWH[+def] manj I

T VP

AUX

V DPi

saw

D DP

himj[+foc, +def]

D NP

OPWH[+def] manj

There is one critical problem with this derivation, namely, Relative Deletion,

which should apply to both the pronoun and to the internal head, cannot apply to the

pronoun. Elements under focus cannot be deleted, because focus cannot be recovered

(Takahashi and Fox 2005:230). Because there are two competing rules (namely Relative

Deletion versus No Deletion of Focussed Material), the derivation is deviant45

. As

expected, the sentence is ungrammatical in Slovene regardless of whether Relative

Deletion applies to the pronoun:

45

As pointed out by Leonard Babby (p.c. 2013), competing rules do not always result in ungrammatical

sentences, as there are resolution strategies that can be applied in some cases. In the case of these two

competeing rules (i.e. Relative Deletion versus No Deletion of Focussed Material), however, it seems no

resolution strategy is available.

81

(62) *Človek, ki sem (njega) videla prihaja.

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC. sawSG.FEM.PAST is-coming3rd.SG.PRES.

'The man that I saw (himLONG-FORM) is coming.'

This accounts both for the obligatory presence of a resumptive pronoun and for its

obligatory clitic morphology in Slovene. This also provides one possible account for the

obligatory absence of any resumptive pronoun in the highest subject position in Slovene

relative clauses. Since Slovene is a subject-pro-drop language, overt subject pronouns are

always [+foc]; there simply are no clitic forms that correspond to a nominative case

pronoun. This means that any overt subject pronoun would be [+foc], and would therfore

automatically raise to the Spec-CP focus positon and thus be deleted under Relative

Deletion.

The absence of a resumptive pronoun in highest subject position at Spell-Out is,

however, a widely-observed phenomenon cross-linguistically. It occurs in all languages

exhibiting resumption, including those whose pronouns are unspecified for focus (e.g.

Hebrew). Therefore, a broader account must be provided.

2.2.2.4. A Universal Condition on Resumptive Pronouns

Among all languages exhibiting canonical resumptive pronouns in relative

clauses, there is a universal restriction prohibiting a resumptive pronoun from occurring

in the highest subject position. This has been explained as relating to the null-subject /

pro-drop parameter, or in more recent terms the Avoid Pronoun Principle, which states

that a pronominal subject can be elided if it is recoverable (see McCloskey 1991, Boeckx

2003). In the case of the highest subject of a relative clause, the pronoun is certainly

82

recoverable, since it must be interpreted as co-referential with the external relative head.

This explanation, however, is not entirely satisfactory, particularly since there are cases in

which resumption is optional (see section 2.2.2.2. for optional resumption in Hebrew and

Chapter 4 for optional resumption in B/C/S and Slovak). If the Avoid Pronoun Principle

were at work, we would not expect to find optional resumptive pronouns at all – they

would either be obligatory or prohibited in any given position.

A possible alternative hypothesis is that all languages function like Slovene, in

that all nominative case pronouns are inherently [+foc], thus forcing them to raise into

Spec-CP and be deleted under Relative Deletion. This would not only prevent resumption

in the highest subject of a relative clause but in any position where the pronoun would

check its nominative case46

. The problem with this analysis, however, is that many

languages do exhibit nominative case [-foc] pronominal subjects, namely expletive

subjects.

The restriction on nominative case resumptive pronouns remains an open question

and a topic for further inquiry.

2.2.3. Overt C0 Relatives versus Wh-Relatives in Slavic

All the examples cited thus far have involved an overt complementizer. In Slavic

languages, the presence of overt C is a necessary condition on the occurrence of a

canonical resumptive pronoun. In this section I will examine the differences between wh-

46

Mitrović (2011) offers the following evidence from B/C/S of a general restriction barring nominative

case resumptive pronouns, even when they do not occupy subject position.

Knjiga, što se (*ona) patljku čita je na stolu.

bookNOM.SG.FEM. that REFL itNOM.SG.FEM. dwarfDAT.SG.MASC. reads is on table.

‘The book that the dwarf feels like reading (*it) is on the table.’

83

relatives and overt-C relatives in Slavic and offer an explanation as to why resumptive

pronouns occur only in the latter. In order to do this, we must take note of three empirical

facts:

i. Resumptive pronouns must co-occur with overt C in a relative

CP.

ii. Overt wh-words cannot occur in the Spec of a CP headed by

overt C.

iii. Overt wh-words cannot occur in the Spec of a relative CP

containing a resumptive pronoun.

2.2.3.1. What happens when an overt wh-word merges?

The stacked-DP structure I have proposed contains two distinct D positions; one

is occupied by the pronoun while the other contains a definite determiner. Because there

are two distinct positions, there is the possibility that both the wh-operator and the

pronoun could be pronounced (this is the case in Macedonian, Greek, and Hebrew). In

many languages (e.g. B/C/S, Slovene, Slovak), however, this is prohibited47

. The stacked

DP shown below is, therefore, a violation of some syntactic rule in those languages.

Diagram 19 (NB: I do not place an asterisk by this structure because it is not universally

prohibited. See footnote 47.)

DP

D DP

pronoun[+def, φ]

D NP

which[+def] R-expression[φ]

47

This restriction is not universal: Greek, for example, allows two case marked D0s to occur within a single

DP, and as this analysis predicts, it also allows wh-relative clauses to contain resumptive pronouns. This

will be addressed in section 2.3.1.2.

84

I propose that this structure is deviant in languages that prohibit case, when assigned to a

DP, from being phonetically realized on more than one D0 in that phrase

48, i.e., case

cannot be realized on both the pronoun and the overt wh-word. But because these

languages require every occurrence of an overt pronoun or of an overt wh-word to be

morphologically marked for case, a stacked-DP containing both is impossible. Thus, as

soon as an overt wh-word is merged into D, a pronoun cannot be merged. Conversely,

when a PF-null wh-operator is merged as the head of DP, a pronoun must then be merged

to that DP (failure to merge the pronoun results in a structure where the case assigned to

DP does not get assigned to any element with PF-legible features, which is automatically

ungrammatical).

48

The notion that the same morphological case cannot be occur on two adjacent D heads is supported by

other independent data from Slovene (data from Marušič and Žaucer 2006):

[DP ta [DP ta zelen svinčnik]]

theNOM.SG.MASC. theNOM.SG.MASC. greenNOM.SG.MASC. pencilNOM.SG.MASC.

'this green pencil'

* [DP tega [DP tega zelenega svinčnika]]

theGEN.SG.MASC. theGEN.SG.MASC. greenGEN.SG.MASC. pencilGEN.SG.MASC.

'of this green pencil'

[DP tega [DP ta zelenega svinčnika]]

theGEN.SG.MASC. theCASE?.SG.MASC. greenGEN.SG.MASC. pencilGEN.SG.MASC.

'of this green pencil'

While these are the data from Marušič and Žaucer 2006, this is not their analysis. They claim that the

second occurrence of ta is not a determiner but rather some other type of head. Given the data that they

present, this ta is at least a defective D0, because it does not inflect for case, number, or gender, and it can

never be stressed.

85

2.2.3.2. What happens when an overt wh-word raises?

There is a strong cross-linguistic generalization that an overt complementizer

cannot cooccur with an overt wh-word in its Spec position49

. This is true not only for

relative clauses (as shown in (64)) but is also for embedded questions, as in (65).

(64) * The book which that I read yesterday was fascinating.

(65) * I wonder which girl that Amy likes.

It has been suggested, following the basic idea of Chomsky and Lasnik's Doubly Filled

COMP Filter (1977), that a deletion rule applies; when an overt wh-word is in Spec-CP,

then an overt C will be deleted. It is unclear precisely when (at what level in the

derivation) and how (by what mechanism) this deletion applies. I suggest that overt C is

PF-deleted according to a condition that prohibits C from being pronounced if an overt

wh-word occupies Spec-CP50

.

2.2.3.3. Getting the Facts

Given the PF-deletion rule in 2.2.3.2 along with the case realization rule in

2.2.3.1, the empirical facts noted in 2.2.3 fall out naturally. Fact (iii) is explained by a

condition on case, namely, the case assigned to DP cannot be phonetically realized on

more that one D0 head in that DP, thus a pronoun D

0 cannot be merged to a DP headed by

an overt wh-word. Fact (ii) is explained by a general restriction on PF: an overt C cannot

be pronounced (i.e. must be deleted at PF) if Spec-CP contains an overt wh-word. Fact (i)

49

It is possible that the Macedonian kojšto (which is a combination of koj 'which' and što 'that') is a

counterexample to this generalization. 50

The precise operation by which the C that gets deleted is immaterial to the current topic, so I will suffice

simply to say that it does delete.

86

is a natural consequence of these two rules taken together: the pronoun can only be

merged into the derivation if there is a PF-null wh-operator. The DP containing the null

wh-operator raises to Spec-CP, thereby stranding the pronoun, but at PF there is no overt

wh-word in Spec-CP and therefore no cause for the deletion of C0. This derives the

consistent co-occurrence of the resumptive pronoun and overt C.

2.2.3.4. The Structure of wh-relatives versus overt C relatives

In many of the Slavic languages that exhibit canonical resumption, the overt-C

and resumptive pronoun strategy is the most common way to form a relative clause when

the site of relativization is the subject51

, direct object or indirect object within the relative

clause. That is, the resumptive pronoun is typically the direct or indirect object of the

relative clause V. Under other conditions, for example (55), when the internal relative

head is the object of a preposition, a wh-relative is typically used.

Slovene

(66) punca, s katero sem plesal

girlNOM.SG.FEM. with whichINST.SG.FEM. AUX1st.SG. dancedSG.MASC.PAST.

'the girl with whom I danced'52

In Slovene, the preposition 'with' takes an instrumental case object. It is this object that is

being relativized in (66). While speakers most often employ the wh-relativization strategy

51

When the site of relativization is the subject of the relative clause, the presence of a resumptive pronoun

in that position is universally prohibited, as mentioned in sections 2.2.2.3 and 2.2.2.4. This could be a result

of the Avoid Pronoun Principle, or alternatively, as I suggest, it could be due to the absence of a [-foc]

nominative case pronoun. 52

English wh-words are sensitive to animacy. 'Who' is used instead of 'which' when referring to an animate

object. Slavic languages do not have a separate wh-word that applies only to animates. So, although I

translate the PP as 'with whom', it is actually literally translated (and glossed) as 'with which'.

87

in such situations, they occasionally use a different relativization strategy, as shown in

(67).

(67)53

punca, ki sem z njo plesal

girlNOM.SG.FEM. that AUX1st.SG. with herINST.SG.FEM. dancedSG.MASC.PAST.

'the girl that I danced with (her).'

In this section, we will compare the derivations of (66) and (67), represented in diagrams

(20) and (21), respectively.

53

Note that the form of the preposition is variable and dependent on PF voicing assimilation. In (66), the

preposition with is s, while in (67) it is z. The preposition is in fact the same; the variation only arises as a

result of regressive voicing assimilation.

88

2.2.3.4.1. The Girl with Whom I Danced (Punca, s katero sem plesal)

(NB: Jaz is the nominative first person singular pronoun and is the elided (pro-dropped)

subject of the relative clause.)

Diagram 20

DP

D NP

Ø

N CP

puncaj

PPi CP

P DP C TP

s ki

D NP DP TP

katero puncoj (jaz)

T VP

sem

V PPi

plesal

P DP

s

D NP

katero puncoj

Since prepositions in Slovene cannot be stranded, the DP katero punco 'which girl', when

it raises into Spec-CP, pied-pipes the preposition s 'with' along with it. The overt

complementizer as well as the internal head are deleted, deriving the PF word order

punca s katero (jaz) sem plesal 'girl with which I danced'.

89

2.2.3.4.2. The Girl that I Danced With Her (Punca, ki sem z njo plesal)54

Diagram 21

DP

D NP

Ø

N CP

puncaj

DPi CP

D NP C TP

OPWH[+def] puncoj ki

DP TP

(jaz)

T VP

sem

V PP

plesal

P DP

s

D DPi

njo[+def]

D NP

OPWH[+def] puncoj

In this derivation, the pronoun njo 'her' is merged to a DP headed by a null wh-operator,

where it checks its instrumental case feature with P s 'with' and checks its definiteness

feature with the internal DP. Once the rest of the items in the relative clause have merged,

the internal DP raises into Spec-CP, leaving behind both the preposition and its clitic

pronominal object.

54

Typically, this structure would be deviant, because prepositions generally take long form pronouns as

their objects. But the term njo, an instrumental case feminine pronoun, is ambiguous between clitic and

long form, so speakers find it possible to use it in a resumptive function.

90

2.2.4. Split Antecedents and Canonical Resumption

As mentioned in section 2.1.3, canonical resumptive pronouns differ from

ordinary pronouns in that they cannot take non-contiguous NPs as their antecedents (i.e.

split antecedents). Recall the difference in grammaticality between the Slovene examples

(43) and (44), repeated here as (68) and (69):

(68) Irinai je povedala Jožetuj, da njenii starši

IrinaNOM.SG.FEM. AUX3rd.SG. toldSG.FEM.PAST. JožeDAT.SG.MASC. that her parentsNOM.PL.

so jimai+j prepovedali živeti skupaj.

AUX3rd.PL. them3rd.DAT.DU. forbidPL.PAST. to-live together

'Irinai told Jožej that heri parents forbid themi+j to live together.'

(69) *Irinai ima-rada človekaj, ki so jimai+j njenii

IrinaNOM.SG.FEM. loves manACC.SG.MASC. that AUX3rd.PL. them3rd.DAT.DU. her

starši prepovedali živeti skupaj.

parentsNOM.PL. forbidPL.PAST. to-live together

'Irinai loves the manj that heri parents forbid themi+j to live together.'

Given the analysis I propose, it is certainly possible for the following structure to exist:

Diagram (22)

DP

D DP

them

D NP

OP[+def]

NP conj NP

Irina and Jože

Here the NP containing both of the antecedents of the pronoun occurs within the

maximal DP headed by the pronoun, which given the Pronominal Reference Condition,

91

forces them to be interpreted as co-referential with Irina and Jože. If the DP Irina and

Jože remains in its first merge position, it will be deleted under Pronominal Associate

Deletion. But if the internal DP is headed by a [+wh] D0, as it would if it were an internal

relative head, then it will raise into Spec-CP.

When the internal DP raises, it must raise as unit, and as a unit, it will have to

match the external relative head. This would result in a sentence such as the following55

:

(70) Jaz sem videla [ Irino in Jožeta ], ki sem ju

I AUX1ST.SG. saw Irina and Jože that AUX1ST.SG. them3rd.ACC.DU.

spoznala na tečaju slovenščine.

met at class SloveneGEN.SG.FEM.

'I saw Irina and Jože, whom I met in Slovene class.'

Thus, while split antecedents are generally permitted (and can be accounted for under my

analysis), they cannot possibly act as antecedents of canonical resumptive pronouns.

2.3. Summary

In this chapter, I have presented a new analysis of relative clauses and a new

account of the phenomenon of canonical resumptive pronouns. I adopt Sauerland's (1998,

2003) model of relative clauses in which there are two heads: one internal to the relative

clause CP and one external. The internal relative head, which takes the external head as

an antecedent, deletes via the operation of Relative Deletion. Building on Sauerland's

model, I propose that cross-linguistically all relative clauses headed by an overt head C0

contain a pronoun in the site of relativization. The pronoun has what I have called a

55

Note that this is a non-restrictive relative clause only because the relative head is composed of proper

names.

92

stacked-DP structure, repeated in Diagram 23. In relative clauses, the internal D must be

an unpronounced wh-operator, thus forcing it to raise to Spec-CP to check an

uninterpretable [wh] feature on C.

Diagram 23

DP

D DP

pronoun[+def]

D NP

OPWH[+def] R-expression

I have also suggested two linguistic parameters. The first parameter is that in some

languages, the internal DP cannot move independently: it obligatorily pied-pipes the

pronoun. In other languages, the pronoun can be stranded, thus yielding a resumptive

pronoun. In languages that have clitic pronouns, the resumptive pronoun is invariably the

clitic form. In languages that do not have two sets of pronouns, we find optional

resumption.

I then examine the difference between overt-C relatives and wh-relatives. I

suggest a second parametric difference, namely, that in some languages in which R-

expressions and articles inflect for case, no more than one D0 can have phonetically

realized case within a single case-marked DP. In such languages, a pronoun cannot be

merged to head a stacked-DP in which the internal D0 is an overt wh-word; when an overt

wh-word is merged, the pronoun cannot then merge without causing a case assignment

violation. Thus, case inflected wh-words and canonical resumptive pronouns occur in

complementary distribution in these languages.

93

2.3.1. Two Predictions

The first prediction that this analysis of canonical resumptive pronouns makes is

that all languages that have clitic pronominal forms (that contrast with long forms56

) also

have canonical resumptive pronouns. In other words, if a language has both long form

pronouns as well as clitic pronouns, the clitic pronouns will occur as canonical

resumptive pronouns in relative clauses. I have looked at many languages with clitic

pronouns, and although I cannot claim that my study has been exhaustive, it has included

all of the Slavic languages, and I have yet to find a counter-example57

.

The second prediction has to do with the two parameters I describe in this chapter.

The stacked-DP analysis presented here includes two fundamental linguistic parameters:

1) some languages allow the internal DP to move independently of the pronoun and 2)

some languages that show morphological case on D0 heads will restrict phonetically

realized case to a single D0. Given any two parameters, we expect four resulting language

types. That is, each cell of the matrix in Diagram 24 should contain at least one element.

Diagram 24

Internal DP: mobile Internal DP: immobile

No case-marking

on >1 D0

Slovene, B/C/S

No rule about case-

marking D0s

English

56

While Romance languages do have clitic pronouns for objects, those pronouns do not contrast with

specific long forms. Slovene, for example, has ga and njega, which are both [ACC.SG.MASC.] pronouns

that differ only in the value of their focus feature. Romance languages do not have these contrasting pairs.

In Spanish, the [ACC.SG.MASC.] pronoun is lo; there is no alternate [+foc] form. 57

Cedric Boeckx (p.c. 2012) also knows of no counter-example to this claim.

94

In the following sections, I will show that the two empty cells of this matrix are filled by

Hungarian and Greek.

2.3.1.1. Hungarian58

Hungarian, like Greek, has stacked-DPs that do not involve pronouns; they have

demonstrative pronouns followed by the definite article.

(71) a jó diák

the good student

'the good student'

(72) ez a jó diák

this the good student

'this good student'

(73) *ez jó diák

this good student

'this good student'

Unlike in Greek, however, the internal DP is immobile, i.e., it cannot move independently

of the stacked-DP:

(74) * a jó diák ez

the good student this

'this good student'

58

Many thanks to E. Wayles Browne and Márton Dornbach for the Hungarian data and grammaticality

judgements.

95

My analysis predicts that Hungarian should not have resumptive pronouns, and indeed,

this is the case. It does, however, have case inflection on determiners. Thus, Hungarian

only has one strategy for forming relative clauses, namely the inflected wh-relativizer.

(75) a lány, akinek elkuldtem a levelet

the girl to-whom I-sent the letter

'the girl to whom I sent the letter'

The table now has three of the four predicted language types.

Diagram 25

Internal DP: mobile Internal DP: immobile

No case-marking

on >1 D0

Slovene, B/C/S Hungarian

No rule about case-

marking D0s

English

In the next section, I show that the final cell contains Greek.

2.3.1.2. Greek

As shown in section 2.2.1.2, morphological case in Greek can occur on more than

determiner in a DP and the internal DP is able to move independently of the stacked-DP.

Thus we would predict that wh-phrases can co-occur with resumptive pronouns, and

indeed this is what we find: (Data from Kotzoglou and Varlakosta 2005)

96

(76) o kokalis proselave ton proponiti ton opio poli tha ton ithelan

the Kokalis hired the coach the whom many would him want

stin omadha tus.

at-the team their.

'Kokalis hired the coach whom many would want (him) on their team.'

Diagram 26

Internal DP: mobile Internal DP: immobile

No case-marking

on >1 D0

Slovene, B/C/S Hungarian

No rule about case-

marking D0s

Greek English

In the next chapter, I will show that Hebrew and Macedonian are in some ways like

Greek, in that multiple D0s from a stacked DP may be pronounced. However they differ

from Greek in that neither of them has overt case marking on articles or R-expression. In

chapter 3, I will examine two non-canonical resumptive pronominal structures: one that

occurs in Hebrew, namely resumptive pronouns left by interrogative wh-movement and

another that occurs in Macedonian, namely object clitic doubling.

97

CHAPTER 3

A BRIEF EXAMINATION OF TWO INSTANCES OF NON-CANONICAL

RESUMPTION

In this chapter we will look at two varieties of non-canonical resumption, which

is defined as the presence of resumptive pronouns (i.e. pronouns that are stranded by the

movement of the internal DP) that are not stranded as a result of wh-operator movement

to Spec-CP of a relative clause. Non-canonical resumptive pronouns are either stranded

by a non-wh movement operation or by wh-movement in some context other than a

relative clause. I will show how the stacked-DP analysis accounts for instances of non-

canonical resumption in Hebrew and Macedonian, and I will expand on the crucial

relationship between personal pronouns and definiteness.

3.1. Hebrew Interrogatives

Resumptive pronouns in Hebrew can occur as a result of interrogative wh-

movement, which is similar to the movement found in relative clauses: both are instances

of movement of a [+wh] element into Spec-CP . While many languages exhibit canonical

resumption, only a few of them also exhibit resumption in interrogatives. The reason for

this is that many languages do not allow the co-ocurrence of an overt wh-element and a

resumptive pronoun (see chapter 2). I claim in chapter 2 that this has to do with a

parametric restriction on case, i.e., case cannot be pronounced on more than one D0

within the DP to which case is assigned. In languages that do not typically show case on

non-pronominal determiners, we would not expect this restriction to apply. In Hebrew, a

98

language that does not mark case on demonstratives, articles, or R-expressions, we

indeed find that, like in Greek, a wh-word may co-occur with a resumptive pronoun.

(77) eyze student nifgaSta i-to?

which student you-met with-him

'Which student did you meet with?'

The structure of (77) is shown in diagram (27):

Diagram 27

CP

DPi CP

D NP C TP

which studentj

DP TP

(you)

T VP

V PP

met

P DP

with

D DPi

himj

D NP

which studentj

Because Hebrew has no overt cases and thus does not have the case restriction found in

Slovene and other languages, a DP headed by an overt wh-word can merge with a

pronominal D0, thereby creating a stacked-DP. When a C

0 with a [wh] feature that needs

to be checked is merged, one of two things can happen. Either the entire stacked DP

raises into Spec-CP, or only the internal DP which student raises, stranding the pronoun.

99

The binding criteria are also met: the pronoun is interpreted as co-referential with the NP

student, which does not c-command the pronoun.

This type of structure is not unique to Hebrew. Boeckx (2003), whose analysis is

similar to the one presented here, points out that pronoun stranding due to interrogative

wh-movement occurs in a wide range of languages (e.g. Irish, Albanian, Romanian, and

Archaic Dutch, among many others). Moreover, for all of the languages in which a

resumptive pronoun is stranded by a wh-phrase, one very strong generalization can be

made: only a D-linked wh-phrase can strand a pronoun. (The data shown for Hebrew

(Sharvit 1999) are also representative of the data observed in these other languages that

strand pronouns due to interrogative wh-movement.)

(78) *Mi nigaSta ito?

who you-met him

'Who did you meet?'

A non-D-linked wh-word cannot strand a pronoun, which is a natural consequence of the

analysis of resumptive pronouns presented in chapter 2. A pronoun must check its

definiteness feature with that of a [+def] head. Moreover, the pronoun must match the

embedded N head in reference. Since neither of these two requirements on the pronoun

can be met in (78), the structure in diagram (28) must be ungrammatical.

Diagram 28

*DP

D NP

pro[+def] who[-def]

100

Since the NP who is indefinite, the pronoun, which must be definite, cannot check its

definiteness feature. The second requirement also cannot be met, because the NP who is

unspecified for φ-features in many languages and is unspecified for reference in all

languages. Thus the pronoun is unable to check its φ-features, and at LF, the Pronominal

Reference Condition (see (45)) cannot be met. The resulting derviation is ungrammatical.

A D-linked wh-word, on the other hand, is both definite59

and has an NP complement,

allowing the pronoun to check its definiteness feature as well as its reference:

Diagram 29

DP

D DP[+def]

pro[+def]φ

D NP

which[+def] studentφ

3.1.1. An Aside on English

English is perhaps unique in having two distinct lexical entries for who, just as it

has two distinct lexical entries for the pronoun he: one is valued positively for

definiteness and the other is valued negatively. The interrogative who is indefinite, but

who can also occur as a relativizer: the animate form of which.

59

Boeckx (2003) points out that in many languages (such as Albanian and Portuguese), D-linked wh-words

are obligatorily marked by a definite determiner.

Cil-ët libra (i) solli Ana?

Which-the books them bought Ana?

'Which books did Ana buy (them)?' Kalluli (1999)

101

(79) Who ate my cake? (interrogative NP who, indefinite)

(80) ?/* The girl which ate my cake is in a lot of trouble. (definite inanimate wh-

relativizer)

(81) The girl who ate my cake is in a lot of trouble. (definite animate wh-relativizer)

In (81), who is a D head, essentially the same as which, however, English has an

additional agreement requirement, according to which a D0 must match its NP sister in

animacy in order for Relative Deletion to apply to the NP. In other words, an NP cannot

be deleted (presumably due to issues of recoverability) if its sister D0 is inanimate.

Diagram 30

NP

N CP

girl

DP CP

...

D NP

which[-anim.] girl[+anim.]

Since which is specified as inanimate in English, it does not match the internal relative

head for animacy in (80), and thus Relative Deletion cannot apply and the derivation

cannot proceed60

.

Slavic languages are quite different from English in this respect. Any instance of

the lexical item who in Slavic languages is specified as indefinite. For this reason, who

does not occur in ordinary relative clauses in Slavic, only in free relatives and

correlatives. This will be discussed further in chapter 5.

60

In interrogatives, Relative Deletion does not apply, so there is no requirement on matching animacy. For

this reason, (80) is ruled out, while the following sentence is not:

Which girl ate my cake?

102

3.2. Macedonian Object Clitic Doubling

Macedonian is one of several languages (others include Albanian and Bulgarian)

that exhibits object clitic doubling. Clitic doubling describes a phenomenon in which a

sentence having a direct or indirect object must additionally include a clitic pronoun

matching that object in φ-features and Case.

Macedonian61

(82) *(Go) vidam Ognen.

himACC.SG.MASC. I-see1st.SG.PRES. OgnenACC.SG.MASC.

'I see Ognen.'

(83) *(Ja) pročitav kniga-ta.

herACC.SG.FEM. I-read1st.SG.AOR. book-theDEF. ACC.SG.FEM.

'I read the book.'

We see from (82) and (83) that a definite object must be doubled by a clitic pronoun. I

suggest that the clitic in clitic-doubling is a resumptive pronoun; it is stranded by the

movement of its sister DP62

. This analysis is quite similar to the analysis of Bulgarian

clitic doubling proposed by Franks and Rudin (2005) in which the clitic is the head of a

K(ase)P and its definite DP associate raises, thereby stranding the pronoun. However,

Franks and Rudin take the position that clitic pronouns (particularly in this function) are

fundamentally distinct from ordinary pronouns, while I maintain that these clitics are

61

Many thanks to Ognen Vangelov for the data and judgments shown for Macedonian. 62

Given this proposal, a reasonable question is whether clitic-doubles can take split antecedents. As shown

in section 2.2.4, canonical resumptive pronouns cannot take split antecedents, which the analysis predicts.

It would also predict that clitic doubling cannot occur with split antecedents, and indeed, the prediction

holds:

* Jas imi+j gix+y dadov ovaa knigax na Ogneni i prativ onaa knigay na Biljanaj.

I themDAT.PL. themACC.PL. gave this book to Ognen and sent that book to Biljana.

'I gave this book to Ognen and sent that book to Biljana.'

103

pronominal. The differences between clitic and long-form pronouns are essentially

prosodic in nature. Consider diagram (31), which reflects the Spell-Out structure of

(83)63

:

Diagram 31

TP

DP TP

(I)

T VP

V DP[-foc]

read

DPi[+def] DP[-foc]

D NP D[+def, -foc] DPi[+def]

the bookj itj

D NP

the bookj

The NP book merges with a definite article, creating the DP the book. The clitic pronoun

D0 then merges to this DP, thereby forming the stacked-DP. The internal DP then raises to

the specifier position of the stacked DP, thus stranding the pronoun.

This derivation invites two questions. The first of these has to do with the

movement leading to resumption. The movement of the internal DP into Spec-DP is

crucial to this analysis and the natural question is why the internal DP should raise to this

position. In fact, this movement does not always occur before Spell-Out. Sportiche

(1992) proposed the Clitic Criterion, repeated below (Note that [+F] represents a

positively valued focus feature):

63

The Macedonian definite article is an enclitic. In this derivation -ta occurs in D0 while its N host is in a

lower position in NP. For this, I adopt the analysis of PF clitic lowering, in which an enclitic will move to a

lower position in order to attach to its host. (Halpern 1992 and Izvorski, King, and Rudin 1997).

104

Clitic Criterion (Sportiche 1992:25)

At LF:

i. A clitic must be in a Spec-Head relationship with a [+F] XP.

ii. A [+F] XP must be in a Spec-Head relationship with a clitic.

When the pronoun is [-foc], the movement of its sister DP into Spec-position may occur

either overtly or as covert LF movement. In many languages, such a movement is strictly

a last resort LF operation. However, this movement may be overt - it is simply a

requirement that the movement occur by LF. Thus, in some languages, the movement to

Spec of the clitic DP may occur within the narrow syntax, as in derivation (31), thereby

giving rise to languages with clitic-doubling64

. But even in these languages, this

movement only occurs prior to Spell-Out if the internal DP is unrecoverable through

discourse reference. Otherwise, the movement will occur at LF.

In diagram (32), I show a similar sentence but one in which the movement of the

internal DP to the Spec of the stacked DP is covert.

Diagram 32

TP

DP TP

(I)

T VP

V DP[-def]

read

D[+def, -foc] DP[+def]

itj

D[+def] NP

the bookj

64

This is similar to the analysis of clitic doubling in Gierling (1997).

105

Here the internal DP does not raise until after Spell-Out at LF. Thus, at PF, the internal

DP is still a sister to the pronoun, and therefore it is deleted under Pronominal Associate

Deletion. This gives the grammatical Macedonian sentence:

(84) Ja pročitav.

herACC.SG.FEM. I-read1st.SG.AOR.

'I read it.'

The second question that the derivation in (31) raises is why the pronominal

double must be a clitic form and not a long form. To answer this, we must first look at the

data. We see from (85) that an indefinite object cannot be doubled. Only definite DPs

may be doubled. This fact falls out naturally from the analysis in which pronouns must

check their definiteness against a definite DP.

However, along with clitic doubling of definite R-expressions, we find that a

[+foc] (i.e. long form) pronoun must also be doubled. In (86), nego, takes contrastive

focus. In order to convey the meaning 'I saw him' without focus on 'him', a [-foc] pronoun

will occur on its own, as in (87).

(85) (*Ja) pročitav (edna) kniga.

herACC.SG.FEM. I-read1st.SG.AOR. someINDEF. ACC.SG.FEM. bookINDEF. ACC.SG.FEM.

'I read a book.'

(86) (Jas) nego *(go) vidov.

I1st.NOM.SG. himACC.SG.MASC. himACC.SG.MASC. I-saw1st.SG.AOR.

'I saw him.'

106

(87) Go vidov.

himACC.SG.MASC. I-saw1st.SG.AOR.

'I saw him.'

The descriptive generalization, then, is that accusative and dative case can only be

checked on one of two phrase types: either an NP (i.e. an indefinite R-expression) or a

DP[-foc] headed by a pronoun. It is unclear precisely why there should be this restriction; it

is a question I leave open. Nevertheless, the analysis I have presented is able to capture

the data. The derivation of (86) according to this analysis is presented in diagram (33)65

:

Diagram 33

TP

DP TP

(Jas)

T FocP

DP[+foc]i FocP

D[+foc] NP Foc VP

negoj Ognenj [uFoc]

V DP[-foc]

vidov

DP[+foc]i DP[-foc]

D[+foc] NP D[-foc] DP[+foc]i

negoj Ognenj go

D[+foc] NP

negoj Ognenj

65

For these derivations, I use the original Macedonian words in order to distinguish clearly between the

long form pronoun and the clitic form.

107

The [+foc] pronoun nego ('him'LONG-FORM) merges with its referent NP, Ognen. Because

nego is not a clitic form, the Clitic Criterion does not apply, and the NP Ognen need not

raise from its merge position as sister to the nego. This DP then merges with the [-foc]

clitic form go ('him'CLITIC), thus creating the stacked-DP headed by [-foc] D0. The internal

DP (containing the NP Ognen and the pronoun nego) is [+foc] and therefore must be

moved into the focus position. Thus it raises successive-cyclically, first into Spec of the

stacked-DP, where it satisfies the Clitic Criterion. The V vidov ('saw') merges and can

check accusative case on its sister DP, since it is headed by a [-foc] pronoun. After V

merges, a [uFoc] head of the focus projection merges66

. The DP headed by the [+foc]

pronoun nego raises once more and lands in a focus position to check the uninterpretable

focus feature heading the focus projection. At PF, the clitic pronoun go then raises to its

fronted position through a linearization rule and the NP Ognen is deleted under

Pronominal Associate Deletion.

3.3. Summary

In this brief chapter, I have introduced two non-canonical instances of

resumption: one resulting from interrogative wh-movement and the other from overt

focus movement (either to check an uninterpretable [+foc] feature or to satisfy the Clitic

Criterion). There are still many questions left open regarding these phenomena,

particularly clitic-doubling, which arises in many different languages and with many

possible cross-linguistically varying limitations. While I do not define these parameters

66

It is unclear whether the FocP is a projection higher than VP or internal to VP, but its relative position to

VP is immaterial to the current analysis.

108

here, I provide an skeletal analysis of stranding in the contexts of clitic-doubling and

interrogative wh-movement on which analyses of related structures may be developed.

The primary analytical tools are the definiteness feature, focus feature, and the

Pronominal Associate Deletion operation. In addition, I invoke Sportiche's (1992) Clitic

Criterion in order to explain the phenomenon of clitic-doubling. The question remains as

to the nature of the specific restriction in Macedonian that forbids accusative and dative

case from being assigned to DP[+foc] headed by a pronoun. Moreover, clitic doubling

differs greatly from language to language; that is, there are a number of language-specific

restrictions that will allow clitic doubling to occur only under specific conditions. I leave

these issues aside for future research.

109

CHAPTER 4

OPTIONAL CANONICAL RESUMPTION IN B/C/S AND SLOVAK

4.1. Introduction

In much of the literature on B/C/S resumptive pronouns (Browne 1986, Goodluck

and Stojanović1996), it has been shown that canonical resumptive pronouns can be elided

under certain conditions. Browne (1986) pointed out that the optionality related to the

animacy of the pronoun's associate; i.e., when the relative head is animate, the resumptive

pronoun is obligatory, whereas when the relative head is inanimate, the pronoun may be

elided. Example (1) is repeated below as (88).

(88) Čovjek što sam *(ga) video.

manNOM.SG.MASC.ANIM. that AUX1st.SG. himACC.SG.MASC.ANIM. sawMASC.SG.PAST.

sjedi ovdje.

sits3rdSG.PRES. here

'The man that I saw is sitting here.'

(89) Film što sam (ga) gledao.

movieNOM.SG.MASC.INAN. that AUX1st.SG. itACC.SG.MASC.INAN. watchedMASC.SG.PAST.

'the movie that I watched'

Bošković (2009) further points out that the gender feature of the relative head plays a

role; when the relative head is feminine, the resumptive pronoun is obligatory, regardless

of its animacy:

110

(90) Kniga, što sam *(je) pročitao

bookNOM.SG.FEM.INAN. that AUX1st.SG. itACC.SG.FEM.INAN. readMASC.SG.PAST.

'the book that I read'

While the animacy of masculine nouns does play a more salient role in the Slavic

languages than the animacy of feminine nouns, it appears that even Bošković's

generalization does not capture the full range of data, when it comes to optional

resumption in B/C/S.

Gračanin-Yuksek (2010) shows that both gender and animacy are, in fact,

unrelated to the optionality of the resumptive pronoun because in some cases, even a

masculine animate is optional.

(91) Iva je upoznala čovjeka, što sam

IvaNOM.SG.FEM. AUX3rd.SG. metMASC.SG.PAST. manACC.SG.MASC.ANIM. that AUX1st.SG.

(ga) vidio.

himACC.SG.MASC.ANIM. sawMASC.SG.PAST.

'Iva met the man that I saw.'

The very same facts hold for Slovak: every instance of optional resumption in B/C/S

corresponds to an instance of optional resumption in Slovak and every instance of

obligatory resumption is B/C/S corresponds to an instance of obligatory resumption in

Slovak. This means that the conditions that permit deletion of the canonical resumptive

pronoun are not unique to B/C/S but apply more broadly. In this chapter, we will examine

what these conditions are and how the analysis of canonical resumption presented in

chapter 2 accounts for it.

111

4.2. Morphological Matching Analysis (Gračanin-Yuksek 2010)

The salient contrast so far is the contrast between sentences (88) and (91). The

embedded relative clauses in the two sentences are identical, and thus the difference

between the two cannot stem directly from any idiosyncratic attributes of the relative

clause (i.e. the properties of the embedded predicate). Moreover, in both sentences, the

relative head is formed from the lexeme čovjek ('man'), which enters the derivation with

the same lexical primitives (that is, the φ-features are valued as masculine and animate in

both). However, while the resumptive pronoun is obligatory in the relative clause in (88),

it is optional in the relative clause in (91). Thus gender and animacy cannot play a role in

determining the optionality of the resumptive pronoun. The difference between (88) and

(91) (in terms of the optionality of the pronoun) must be syntactic and not simply a

consequence of the lexical features on the relative head.

The clearest observable difference between (88) and (91) is that the external

relative head in (88) is nominative, while the external relative head in (91) is marked

accusative. Gračanin-Yuksek (2010) claims that this is the crucial difference between the

two. More precisely, she claims (adopting the Sauerland matching model of relative

clauses) that it is by means of a matching effect of the external and internal relative heads

that the deletion of the internal relative head is licensed. That is, if the external relative

head and the internal relative head are identical, the internal relative head undergoes

deletion. Furthermore, she shows that this matching relation is not based upon the

identity of the features (e.g. abstract case) of the heads but rather the morphological

identity of the two heads, and thus, the matching effect can obtain between two featurally

distinct but syncretic relative heads.

112

(92) Dijete [ što sam (ga) vidio ] voli

childNOM.SG.NEUT. that AUX1st.SG. itACC.SG.NEUT. sawMASC.SG.PAST. loves3rd.SG.PRES.

Ivu.

IvaACC.SG.FEM.

'The child that I saw loves Iva.'

While dijete (child) is nominative case in the matrix clause, the resumptive pronoun is

accusative. Nevertheless, the resumptive pronoun is optional. The reason that it is

optional, according to Gračanin-Yuksek (2010), is that the form dijete is syncretic

between nominative and accusative.

Gračanin-Yuksek (2010) develops an analysis in which resumptive pronouns are

in free variation with N0 internal relative heads; either a resumptive pronoun may merge

or relative head may merge into the relative clause. She proposes that once a resumptive

pronoun is merged, there can be no matching relation. If, on the other hand, an internal

relative head that matches the external relative head is merged, it can raise and then

delete under Relative Deletion. Thus, the derivation of (92) without a resumptive pronoun

can, according to Gračanin-Yuksek, be represented as follows in Diagram (34). (Note: the

subject of the relative clause is the unpronounced term Ja, which is the 1st person

singular pronoun, I.)

113

Diagram 34

TP

DP TP

D NP T VP

Ø

N CP V DP

Dijete voli Ivu

DP CP

dijete

C TP

što

DP TP

(Ja)

T VP

sam

V DP

vidio dijete

The word for child in B/C/S is dijete for both nominative and accusative. Thus, the

internal relative head and the external relative head match in form, although they do not

match in case. Gračanin-Yuksek argues that this morphological matching is what

ultimately results in those relative clauses that do not contain resumptive pronouns. This

also accounts for the data in sentence (89) because film, like dijete, is syncretic between

nominative and accusative. Thus the accusative marked internal relative head can raise

and be deleted on morphological identity with the external relative head.

Gračanin-Yuksek also claims that this raising cannot occur with oblique objects.

This is a necessary conjecture, given her analysis, because oblique resumptive pronouns

are never optional, regardless of whether or not there is a morphologically matching

antecedent in the higher clause.

114

(93) Napisala sam pismo čovjeku, što si

wroteFEM.SG.PAST. AUX1st.SG. letterACC.SG.NEUT. manDAT.SG.MASC. that AUX2nd.SG.

*(mu) pomogao.

himDAT.SG.MASC. helpedMASC.SG.PAST.

'I wrote a letter to the man that you helped.'

Assuming Gračanin-Yuskek's analysis, there should be two initial options available in

sentence (93): either the resumptive pronoun could merge or the N0 relative head could

merge. If a relative head merges into the derivation instead of a resumptive pronoun, the

internal relative head would be identical to the external relative head, as in (94).

(94) Napisala sam pismo čovjeku

wroteFEM.SG.PAST. AUX1st.SG. letterACC.SG.NEUT. manDAT.SG.MASC.

[CP što si pomogao (čovjeku). ]

that AUX2nd.SG. helpedMASC.SG.PAST. manDAT.SG.MASC.

'I wrote a letter to the man that you helped.'

If the internal relative head were to raise, it would be deleted under Relative Deletion,

thus generating the ungrammatical sentence (95), in which there is no resumptive

pronoun:

(95) * Napisala sam pismo čovjeku, što si

wroteFEM.SG.PAST. AUX1st.SG. letterACC.SG.NEUT. manDAT.SG.MASC. that AUX2nd.SG.

pomogao.

helpedMASC.SG.PAST.

'I wrote a letter to the man that you helped.'

115

Thus, it is critical that Gračanin-Yuksek include in her analysis some way to rule out

sentences like (95). She claims that oblique heads cannot raise and supports her claim by

showing that in other structures, specifically in Free Relatives, oblique heads cannot

raise. The evidence comes from reconstruction effects in free relatives:

(96) Janj će pohvaliti [RC koje god svojei/*j

JanNOM.SG.MASC. will praise whichACC.SG.NEUT. ever self'sACC.SG.NEUT.

dijete Ivani dovede ].

childACC.SG.NEUT. IvanNOM.SG.MASC. brings

'Janj will praise whichever of hisi/*j children Ivani brings.' (Gračanin-Yuksek 2010)

The reflexive possessive pronoun svoje in (96) must be anaphoric on Ivan and not Jan.

Since the anaphor is not c-commanded by Ivan at PF, it is only through reconstruction of

the anaphor into its first merge position that the correct interpretation can be derived.

Gračanin-Yuksek demonstrates that reconstruction is not possible when the relative

pronoun is oblique.

(97) * Vidi se sjeća [ kojeg god svogi/j

VidNOM.SG.MASC. REFL remembers whichGEN.SG.MASC. ever self'sGEN.SG.MASC.

psa se Janj bojao. ]

dogGEN.SG.MASC. REFL JanNOM.SG.MASC. feared

'Vid remembers whichever of his dogs Jan feared.'

She claims that because oblique arguments cannot reconstruct, they cannot raise, and she

extends this generalization to oblique arguments in relative clauses; i.e., they too are

unable to raise.

116

4.3. Slovak

A major impediment to the Gračanin-Yuksek analysis is that Slovak, a language

that has the same restrictions on optional resumption as B/C/S, does not have the same

restriction on oblique argument reconstruction.

4.3.1. Slovak Data Corresponding to B/C/S Data

Slovak, like B/C/S, frequently requires a resumptive pronoun in relative clause

CPs headed by an over C0. In (98), which corresponds to the B/C/S sentence in (88), the

resumptive pronoun obligatory.

(98) Človek, čo som *(ho) videl sedí

manNOM.SG.MASC. that AUX1st.SG. himACC.SG.MASC. sawMASC.SG.PAST sits3rdSG.PRES.

tu.

here.

'The man that I saw is sitting here.'

Like speakers of B/C/S, speakers of Slovak often have the intuition that the pronoun in

(98) is obligatory due to the animacy of the relative head. However, as illustrated by (99),

this is not an adequate explanation.

(99) Kúpil som si knihu od človeka

boughtMASC.SG.PAST. AUX1st.SG. REFL bookACC.SG.FEM. from manGEN.SG.MASC.

čo si (ho) stretol.

that AUX2nd.SG. himACC.SG.MASC. metMASC.SG.PAST.

'I bought myself a book from the man that you met.'

117

The resumptive pronoun in (99) is not obligatory, although the relative head is animate.

Moreover, the external relative head and the resumptive pronoun are assigned two

distinct cases. While the external relative head človeka 'man' has its genitive case checked

by the preposition od 'from', the resumptive pronoun ho 'him' checks its structural

accusative case as the object of stretol 'met'.

In Slovak, as in most Slavic languages, animate masculine nouns are syncretic

between genitive and accusative forms. Thus the form of the internal relative head would

be identical to the form of the external relative head. Given Gračanin-Yuksek's analysis,

the derivation would be as follows:

(100) Kúpil som si knihu od človeka

boughtMASC.SG.PAST. AUX1st.SG. REFL bookACC.SG.FEM. from manGEN.SG.MASC.

[CP človeka [CP čo si človeka stretol. ]]

manACC.SG.MASC. that AUX2nd.SG. manACC.SG.MASC. metMASC.SG.PAST.

'I bought myself a book from the man that you met.'

In (100), the relative head (not the resumptive pronoun) is merged into the derivation. It

raises to Spec-CP where it deletes on morphological identity with the external relative

head. This corresponds to the B/C/S sentence (92) in which dijete 'child' can be deleted

on identity with the external relative head.

Slovak is also like B/C/S insofar as oblique resumptive pronouns are uniformly

obligatory, regardless of whether or not they match their external relative heads.

118

(101) Napísala som list človekovi, čo si

wroteFEM.SG.PAST. AUX1st.SG. letterACC.SG.MASC. manDAT.SG.MASC. that AUX2nd.SG.

*(mu) pomohol.

himDAT.SG.MASC. helpedMASC.SG.PAST.

'I wrote a letter to the man that you helped.'

The external relative head in (101), človekovi 'man', is dative, being the indirect object of

V napísala 'wrote'. The resumptive pronoun is direct object of the verb pomohol 'helped',

which is a predicate that checks quirky dative case on its object. Thus, the resumptive

pronoun and the external relative head are both dative.

If instead of the resumptive pronoun, the R-expression relative head were merged

into the relative clause, it would match (both morphosyntactically and morphologically)

the external relative head: both would be realized as človekovi (manDAT.SG.MASC.).

According to Gračanin-Yuksek, the problem with merging this oblique internal

relative head is that it cannot raise (because it cannot reconstruct) and undergo Relative

Deletion. Thus, the derivation is deviant and the only option is to merge a resumptive

pronoun. But the only independent evidence she provides for arguing that oblique objects

do not raise comes from raising in free relatives.

4.3.2. Slovak Free Relatives

While the Slovak data patterns with the B/C/S data with respect to canonical

resumptive pronouns, Slovak does not have a restriction on reconstructing oblique objects

in free relatives67

.

67

Some speakers consider the B/C/S analog of (102) to be grammatical. Those who do find it grammatical

also affirm that the reflexive possessive pronoun can only be anaphoric on the relative clause internal NP.

119

(102)Jožoj pomôže ktorýmkoľvek svojími/*j

JožoNOM.SG.MASC. will-help whicheverDAT.PL.MASC. self'sDAT.PL.MASC.

priateľom Ivani poslal listy.

friendsDAT.PL.MASC. IvanNOM.SG.MASC. sentMASC.SG.PAST. lettersACC.PL.MASC.

'Jožoj will help whichever of hisi/*j friends Ivani sent letters to.'

The relative clause in (102) contains the reflexive possessive pronoun svojím, which must

be interpreted as anaphoric on the lower NP Ivan and cannot be interpreted as anaphoric

on the matrix subject NP Jožo. Thus the dative case marked phrase ktorýmkoľvek svojím

priateľom ('whichever of his friends') must reconstruct into its first merge position where

it is c-commanded by Ivan.

Because Slovak allows raising of oblique objects, Gračanin-Yuksek's analysis

does not hold, i.e., given her analysis, there is no compelling reason why matching and

relative deletion of oblique objects cannot occur in Slovak. Thus, there is no way for

Gračanin-Yuksek's analysis to rule out the following ungrammatical sentence:

Bogdanj će pomoći kojim god je svojimi/*j prijateljima

BogdanNOM.SG.MASC.will help whichDAT.PL.MASC. ever AUX3rd.SG. self'sDAT.PL.MASC. friendsDAT.PL.MASC.

Ivani poslal pisma.

IvanNOM.SG.MASC. sentMASC.SG.PAST. lettersACC.PL.NEUT.

'Bogdanj will help whichever of hisi/*j friends Ivani sent letters to.'

Because some speakers do allow for reconstruction of oblique objects even in B/C/S, it is not a

strong argument for a general restriction against raising oblique arguments. Many thanks to Muamera

Bregović for her insights.

120

(103) *Napísala som list človekovi, čo si

wroteFEM.SG.PAST. AUX1st.SG. letterACC.SG.MASC. manDAT.SG.MASC. that AUX2nd.SG.

pomohol.

helpedMASC.SG.PAST.

'I wrote a letter to the man that you helped.'

4.4. Optional Resumption under the Stacked-DP analysis

The stacked-DP analysis has some elements in common with Gračanin-Yuksek's

analysis: both incorporate the Sauerland matching analysis of relative clauses, in which

there are two distinct heads. The crucial difference between the two analyses is that in

Gračanin-Yuksek (2010), the resumptive pronoun and the N0 internal relative head occur

in free variation. Under the stacked-DP analysis, all relative clauses contain both a

pronoun and its associate, which is the internal relative head. In addition, the Gračanin-

Yuksek analysis relies on the inertness of oblique objects in order to explain the

obligatory presence of oblique case marked resumptive pronouns. In the previous section,

I illustrated how her argument that oblique objects are inert is problematic.

4.4.1. Optional Non-Oblique Resumptive Pronouns

Gračanin-Yuksek's most striking observation is that it is the form of the relative

head that determines whether the resumptive pronoun is optional. Even if the forms are

merely syncretic and bear distinct morphosyntactic features (particularly abstract case),

the resumptive pronoun is nevertheless optional. This observation indicates that the

121

mechanism employed to delete the resumptive pronoun operates at the level of PF, where

abstract case is illegible and only morphological form is relevant.

Given the stacked-DP analysis of resumptive pronouns, the relative clause in

(100) (repeated below) would be derived as follows:

(100) Kúpil som si knihu od človeka

boughtMASC.SG.PAST. AUX1st.SG. REFL bookACC.SG.FEM. from manGEN.SG.MASC.

[CP človeka [CP čo si človeka stretol. ]]

manACC.SG.MASC. that AUX2nd.SG. manACC.SG.MASC. metMASC.SG.PAST.

'I bought myself a book from the man that you met.'

Diagram 35

PP

P DP

od

D NP

Ø

NGEN CP

človeka

DP CP

D NPACC C TP

OPWH človeka čo

DP TP

(ty)

T VP

si

V DPACC

stretol

D DPACC

ho

D NPACC

OPWH človeka

122

In this model of relative clauses and this analysis of resumptive pronouns, the PF

operation that allows optional deletion must apply only when the internal head and the

external head are a complete match. However, the raised copy of the internal relative

head (the copy in Spec-CP) undergoes relative deletion before Spell-Out. In other words,

while Relative Deletion occurs in narrow syntax, the ellipsis of the resumptive pronoun

occurs at PF. Because the moved internal relative head (i.e. the highest copy of the

internal relative head) is deleted prior to Spell-Out, it is not relevant at PF. Therefore, the

PF matching relation is not between the the external head and the highest copy of the

internal relative head, but rather the matching relation obtains between the external head

and the first merge (i.e. lowest) copy of the relative head.68

The PF deletion of the resumptive pronoun operates as follows: if the external

head and the internal head match, the pronoun with which the internal head is associated

may optionally be deleted. The mechanics of this operation are fairly simple: if there

exists a matching relation between the external and internal relative heads, the DP

containing the internal relative head is deleted.

However, we see in diagram (35) that the lowest copy of the internal relative head

is part of a stacked DP. Thus, given the matching relation between the external and

internal relative heads, either the minimal DP containing the internal relative head may be

deleted or, alternatively, the maximal DP containing the internal relative head may be

68

This is not to say that the derivational relationship between the two positions of the internal relative head

(i.e. first merge and Specifier of the relative CP) is relevant at PF. The Spec-CP of the relative clause is

deleted (via Relative Deletion) prior to Spell-Out, so it is, in fact, invisible to PF. It is only the first merge

copy that is relevant at PF and only for the deletion of the resumptive pronoun under morphological

matching of the external and internal relative heads.

123

deleted. In the former case, the resumptive pronoun will be pronounced, while in the

latter case, it will be elided.

4.4.2. Obligatory Oblique Resumptive Pronouns

Bianchi (2004) argues that while structural case does not necessarily need to be

spelled out, inherent case must be spelled out. She offers several pieces of empirical

evidence supporting this claim. For example, she cites a phenomenon in Russian

involving the genitive of quantification (Babby 1987).

(104) Ja kupila [QP pjat' knig. ]ACC

INOM.SG. boughtFEM.PAST. iveACC. booksGEN.PL.FEM.

'I bought five books.'

In (104), the phrase pjat' knig ' five books' is a Quantifier Phrase assigned structural

accusative case as the direct object of V kupila 'bought'. While the quantifier pjat' 'five'

surfaces in the accusative case, the quantified object knig 'books' is assigned genitive by

the quantifier. Now see (105).

(105) My razgovaryvali o [QP pjati knigax ]LOC

WeNOM.PL. were-talkingPL.PAST. about fiveLOC. booksLOC.PL.FEM.

'We talked about five books.'

In (105), however, the QP is assigned inherent locative case by the preposition o 'about'.

The quantifier cannot assign genitive to the quantified object, because in instances of case

conflict, inherent case is favored over structural case (Babby 1987). Thus, a quantified

NP will surface as genitive, so long as the QP is assigned structural case, whereas if the

124

QP is assigned inherent case, the genitive of quantification is blocked and inherent case is

pronounced on the NP (Babby 1987).

Because oblique case must be pronounced, oblique case-marked resumptive

pronouns cannot be deleted regardless of matching effects. In terms of the structure, this

means that even when the external relative head and the internal relative head match at

PF, only the minimal DP containing the internal relative head can be elided if the

maximal DP headed by the resumptive pronoun is assigned oblique case.

4.5. Summary

This chapter concerns the phenomenon of optional canonical resumptive

pronouns in B/C/S and Slovak. Both languages exhibit the same patterns in their data

with respect to licensing the deletion of resumptive pronouns. As pointed out by

Gračanin-Yuksek (2010), the omission of a resumptive pronoun is dependent upon a

morphological matching of the internal and external relative head. While I follow the

basic insight of Gračanin-Yuksek's analysis (the operations involved are matching and

deletion), my analysis is significantly different from hers with regard to how those

operations apply.

Gračanin-Yuksek posits a lexical difference between sentences that contain a

canonical resumptive pronoun and those that do not; that is, she claims that the

numerations for the following two sentences are distinct:

125

(106) Iva je upoznala čovjeka, što sam

IvaNOM.SG.FEM. AUX3rd.SG. metMASC.SG.PAST. manACC.SG.MASC.ANIM. that AUX1st.SG.

ga vidio.

himACC.SG.MASC.ANIM. sawMASC.SG.PAST.

'Iva met the man that I saw.'

(107) Iva je upoznala čovjeka, što

IvaNOM.SG.FEM. AUX3rd.SG. metMASC.SG.PAST. manACC.SG.MASC.ANIM. that

sam vidio

AUX1st.SG. sawMASC.SG.PAST.

'Iva met the man that I saw.'

This difference has no consequences at LF in either B/C/S or in Slovak. In the analysis I

present in section 4.4.1, on the other hand, the difference between (106) and (107) is a

purely PF-level phenomenon. Upon the matching of the internal and external relative

head, a deletion operation will apply to the DP containing the internal relative head. This

operation can either apply to the minimal DP, which will then leave the resumptive

pronoun to be pronounced (thus resulting in sentence (106)), or alternatively, it can apply

to the maximal DP, which will result in the deletion of both the internal relative head as

well as its associated pronoun (thus resulting in sentence (107)).

I also prove that the argument Gračanin-Yuksek offers to explain obligatory

oblique resumptive pronouns is inadequate. I propose instead (following Bianchi 2004)

that resumptive pronouns are obligatory when they are marked with inherent case. This

notion of obligatory pronunciation of inherent case is supported by independent empirical

evidence.

126

CHAPTER 5

A FEATURE MISMACH MYSTERY IN SLOVENE

As shown in the previous chapters, many aspects of resumption vary between

languages and even within a single language. In some languages, such as Hebrew, the use

of canonical resumptive pronouns appears to be optional: their use depends upon whether

or not the stacked-DP or the internal DP raises to Spec-CP. In some languages (e.g.

Macedonian), resumption occurs as a result of the internal DP movement to Specifier of

the stacked-DP, yielding clitic doubles of objects. In short, there is a variety of parameters

that cause resumption to be restricted in different ways in different languages.

Nevertheless, there are some aspects of resumption that are considered to be universal

constants: (1) resumptive pronouns are syncretic with ordinary pronouns, (2) resumptive

pronouns are blocked from highest subject position of a relative clause, (3) resumptive

pronouns are obligatorily anaphoric on a proximate overt antecedent, and (4) resumptive

pronouns share the same φ-features and denotation as their overt antecedent. It is the last

of these generalizations and an apparent counter-example to it (introduced in Section 5.2)

that will be investigated in this chapter.

5.1. A Brief Typology of Slovene Relative Clauses

Slovene, a South Slavic language spoken by just over 2 million people in the

northwest region of former Yugoslavia, has a morphologically rich array of relative

clause types. In this section, we will examine the distinct strategies of forming relative

clauses in Slovene.

127

5.1.1. Kateri-type relatives (which-relatives)

Relative clauses headed by a D0 which relativizer are widely attested cross-

linguistically. In fact, in the analysis I have presented for relative clauses, a wh-relativizer

is always present (either as the overt lexical item which or as a silent wh-operator) in all

derivations of relative clauses. In Slovene, kateri ('which') is declined as an adjective,

inflecting for case, number, and gender. The table below gives the full declension

paradigm of kateri.

Diagram 36

SG SG SG DU DU DU PL PL PL

MASC FEM NEUT MASC FEM NEUT MASC FEM NEUT

NOM Kateri katera katero katera kateri katera kateri katere katera

ACC kateri/

katerega69

katero katero katera kateri katera kateri katere katera

GEN katerega katere katerega katerih katerih katerih katerih katerih katerih

DAT kateremu kateri kateremu katerima katerima katerima katerim katerim katerim

LOC katerem kateri katerem katerih katerih katerih katerih katerih katerih

INST katerim katero katerim katerima katerima katerima katerimi katerimi katerimi

Although kateri exhibits adjectival morphology, it is, in fact, a determiner head. While

adjectives can occur in DPs headed by the [+foc] deictic determiner ta, kateri cannot.

(108) ta velika hiša

this big house

'this big house'

(109) * ta katera hiša

this which house

69

For animate masculine nouns, the accusative morphology follows the genitive declension morphological

pattern.

128

The D0 kateri can either precede an NP in an interrogative, as in (110), or it can introduce

a relative clause, as in (111). In both instances, it is a [+def] determiner.

(110) Katero knjigo si prebral včeraj?

whichACC.FEM.SG. bookACC.FEM.SG. AUX2nd.SG. readMASC.SG.PAST. yesterdayADV.

'Which book did you read yesterday?'

(111) Knjiga, ob kateri sem se nasmejala,

bookNOM.SG.FEM. about whichLOC.SG.FEM. AUX1ST.SG. REFL laughedFEM.SG.PAST.

je Moskva 2042.

is3rd.SG.PRES. Moscow 2042.

'The book that I was laughing about was Moscow 2042.'

It is important to note that the kateri-type relative is almost exclusively used as

the object of a preposition - a position that cannot be occupied by a clitic pronoun (see

footnote 31, chapter 2).

5.1.2. Free Relatives and Co-Relatives70

5.1.2.1. The Morphology of Relative wh-pronouns

In most Slavic languages, wh-relativizers are morphologically identical to their

interrogative wh counterparts, as illustrated by the Slovak examples below.

70

A Free Relative is a relative clause that contains a wh-phrase but does not contain an overt relative head,

as in the example below:

You reap what you sow.

A Co-relative is a construction related to the free relative (it, too, has no overt relative head), but it contains

a demonstrative referring to the wh-pronoun.

You reap that, which you sow.

129

Slovak

(112) Koho tu nepoznáš?

whoACC.SG. here you-don't-know2nd.SG.NEG.PRES.

'Who don't you know here?'

(113) Nedôveruj (tomu), koho

don’t-trust2nd.SG.NEG.IMPERATIVE oneDAT.MASC.SG.DEM.PRON. whoACC.SG.

nepoznáš.

you-don't-know2nd.SG.NEG.PRES.

'Don't trust who you don't know.'

Slovene is different in this respect from the other Slavic languages, since all Slovene

relative wh-pronouns are morphologically marked with an -r affix. They are otherwise

similar, if not identical, to their interrogative counterparts. Listed below are some

examples of interrogative wh-pronouns and their corresponding relative wh-pronouns.

Diagram 37

Interrogative wh-pronoun Relative wh-pronoun

whatNOM./ACC. kaj kar

where kje kjer

when kdaj kadar

how kako kakor

whoNOM. kdo kdor

whoACC./GEN. koga kogar

whomDAT. komu komur

Whose čigav čigar

130

The -r morpheme is a reflex of an early Slavic affix -že, which also created relative wh-

pronouns from interrogative ones. In Slovene and B/C/S, an intervocalic -ž- sometimes

became -r-, thus the relative wh-pronominal affix became -re. Eventually in Slovene, the

final vowel was lost (Greenberg 1999)71

.

5.1.2.2. The Function of Relative wh-pronouns

Unlike English, Slovene employs wh-relatives (with the exception of kateri-

relatives, i.e. which-relatives) in an extremely limited functional capacity. In English, wh-

relatives may often be used interchangeably with that-relatives.72

(114) a. the man that I know

(114) b. the man who(m)/*which I know

(115) a. the book that I read

(115) b. the book which I read

(116) a. the reason that I called

(116) b. the reason why/*which/for which I called

(117) a. the place that we went (to)

(117) b. the place where we went (*to)

(117) c. the place which we went *(to).

71

Many thanks to E. Wayles Browne for bringing to my attention both the evolution of the -r morpheme as

well as the Greenberg (1999) article on its development. 72

English is different from Slovene in that the relative wh-pronouns of English are essentially

amalgamations of a preposition and the word which; i.e., the relative wh-word why is a reduction of the PP

for which. While I have no formal analysis of this, (117a) - (117c) offer evidence that an analysis along

these lines must be correct. While in (117b) the preposition to is impossible, in (117c) it is required. This

suggests that the preposition to is an inherent component in the wh-word where as it is used here. In other

words, the relative wh-word where is, in some sense, a composite of the items to and which.

131

In contrast to the wide application of wh-relatives in English, most Slavic

languages employ wh-relatives (besides which-relatives) exclusively in free and co-

relatives, as in the Slovene examples below (examples (118)-(120) are from

Chidambaram 2007).

Slovene

(118) Kar bo presodilo sodišče, bomo sprejeli.

what will judge/rule court we-will accept.

'Whatever the court will rule, we will accept.' (Free Relative)

(119) Si to, kar ješ.

you-are thatDEM. what you-eat

'You are that which you eat.' (Co-Relative)

(120) Vse, kar potrebujejo, je ljublezen.

everything what they-need is love.

'All (what) they need is love.' (Free Relative)

The wh-relative pronoun kar, shown here, cannot occur in an ordinary (restrictive or non-

restrictive) relative clause.

(121) * To jabolko, kar ješ ni še zrelo.

that apple what you-eat isn't yet ripe

'That apple which you are eating isn't ripe yet.'

132

The reason that (121) is ungrammatical in Slovene is that wh-pronouns are identical to

their interrogative counterparts in their features, including their definiteness feature.

Because wh-pronouns are indefinite (with the crucial exception of which), they cannot

occur in ordinary relative clauses, since ordinary relative clauses require a definite

internal relative head. Instead their use is limited to relative clauses in which the internal

relative head is a bound variable, i.e. free and co-relative clauses. The only way in

Slovene to express the meaning intended in (121) is to use what I will refer to as a ki-type

relative.

5.1.3. ki-type relatives (that-relatives)

In Slovene, most subordinate clauses are headed by C0 da, which translates as

that.

(122) Vem, da ideš.

I-know that you-go.

'I know that you are going.'

But unlike English, in which the subordinating complementizer can also be used to head

relative clauses, the Slovene C0 da can never head a relative clause. Slovene has a

specialized complementizer whose unique function is to head restrictive and non-

restrictive relative clauses73

.

73

Sentence (123) is ambiguous, because the relative clause could either be interpreted as restrictive (if the

listener thinks I may have more than one sister) or as non-restrictive (if the listener is aware that I have only

one sister).

133

(123) Moja sestra, ki živi v Čikagu,

myNOM.SG.FEM. sisterNOM.SG.FEM. that lives3rd.SG.PRES. in ChicagoLOC.SG.NEUT.

je odvetnica.

is3rd.SG.PRES. lawyerNOM.SG.FEM.

'My sister who lives in Chicago is a lawyer.'

'My sister, who lives in Chicago, is a lawyer.'

In the relative clause in (123), there is a gap in the highest subject position of the

relative clause. The canonical subject position, Spec-TP, is occupied by an unpronounced

copy of the internal relative head that raised to Spec-CP. Since Slovene is a subject pro-

drop language, there is never a resumptive pronoun in Spec-TP. However, if the internal

relative head raises from any other position within the ki-type relative clause, there will

be a resumptive pronoun stranded by that movement.

(124) Govorila sem s punco, ki si *(jo)

spokeSG.FEM. AUX1st.SG. with girlINST..SG.FEM. that AUX2nd.SG. herACC.SG.FEM.

spoznala včeraj.

metSG.FEM. yesterday

'I spoke with the girl that you met yesterday.'

The resumptive pronoun in (124) is obligatory. It should also be noted that, while their

Case features are distinct, all of the φ-features (number and gender) of the pronoun and

the external relative head, punco 'girl', match. This comes as no surprise, given both the

Pronominal Reference Condition as well as the general condition on relative clauses that

the external head and the internal head must be co-referential (Sauerland 1998, 2003).

134

(125) Kupila sem ta dva stola, ki

boughtFEM.PAST. AUX1ST.SG. thoseACC.DU.MASC. two chairsACC.DU.MASC. that

ju je prodajala Ana.

themACC.DU.MASC. AUX3rd.SG. soldFEM.PAST. AnaNOM.SG.FEM.

'I bought those two chairs that Ana was selling (them).'

In (125) the relative head is a pair of objects, more precisely a pair of chairs (stol 'chair' is

masculine in Slovene). The resumptive pronoun, therefore, must have both the dual

number feature and the masculine gender feature.

The Slovene ki-type relative, with its indeclinable C0 and resumptive pronoun, is

typologically common among the world's languages. It follows the same basic pattern

occurring in every other language exhibiting canonical resumption: it contains an

indeclinable C0 and a resumptive pronoun whose φ-features and reference match those of

the external relative head but whose case feature is determined by its grammatical

function within the relative clause.

5.2. Slovene Superlative Clauses (kar-phrases)

The clauses that frequently follow comparative adjectives (i.e. than-clauses) and

superlative adjectives (i.e. that-clauses) are often compared to if not equated with relative

clauses (e.g. Andrews 1985, Kennedy 1997, Sauerland 1998). In many languages, such as

English, the clause following a superlative is indistinguishable from a relative clause:

(126) The seeker [that played for Hogwarts] caught the Golden Snitch. (relative clause)

(127) The best seeker [that played for Hogwarts] was Harry Potter. (superlative clause74

)

74

The reasons for distinguishing a superlative clause from a relative clause are identified in section 5.2.2

135

In (126) and (127) the embedded clauses headed by C0 appear identical. The purpose of

the following sections will be to demonstrate, using data from Slovene, that these clauses

are actually structurally distinct.

5.2.1. ki-type relatives versus kar-phrases

In chapter 2, the ki-type relative is shown to contain a canonical resumptive

pronoun; that is, the pronoun must match the external relative head in φ-features and

reference. Any deviation from this will yield an ungrammatical sentence, as shown in

example (128):

(128) *Študenti je videl profesorjaj, ki so

studentNOM.SG.MASC. AUX3rd.SG. sawPAST.SG.MASC. professorACC.SG.MASC. that AUX3rd.PL.

jim*i/*j/*k dali nagrado.

themDAT.PL. gavePAST.PL. awardACC.SG.FEM.

'The student saw the professor that they gave the award to (them).'

In (128), the number feature mismatch between the resumptive pronoun (plural) and the

relative head (singular) causes the sentence to be ungrammatical. Even if there exists a

potential antecedent in the higher clause, if that antecedent is not the external relative

head, the sentence is still ungrammatical.

and 5.2.3.

136

(129) *Študentjei so videli profesorjaj, ki so studentNOM.PL.MASC. AUX3rd.PL. sawPAST.PL.MASC. professorACC.SG.MASC. that AUX3rd.PL.

jim*i/*j/*k dali nagrado. themDAT.PL. gavePAST.PL. awardACC.SG.FEM.

'The students saw the professor that they gave the award to (them).'

The kar-phrase, like a relative clause, is a clausal modifier of a nominal head.

Like a ki-type relative, it is headed by an indeclinable C0 and obligatorily includes a

pronomimal element.

(130) V New-Yorku lahko vidiš najvišje zgradbe, kar so

In New-York can you-see tallestACC.PL.FEM. buildingsACC.PL.FEM. that AUX

jih kdaj zgradili v Ameriki.

themACC.PL.FEM. ever they-built in America

'In New York you can see the tallest buildings that they ever built (them) in

America.'

Unlike the wh-relativizer kar in 5.1.2.1., the word kar in the embedded clause in (130) is

not an inflected form. While it is homophonous with the nominative and accusative case

wh-relativizer, this instance of kar in (130) is not case-marked at all. The form of kar in

these constructions (that is, ordinary relative clauses, as opposed to free or co-relatives) is

independent of case and feature considerations. In sentence (131) below, the case of the

kar-phrase internal head must be dative, since it is the indirect object of the embedded V

dali 'gave'. Nevertheless, the word kar maintains its form irrespective of the features or

case assignment of the internal head. In fact, it cannot be replaced by the dative wh-

relativizer.

137

(131) Tiste so najboljše glasbenice, kar/*komur so

those are the-bestNOM.PL.FEM. musiciansNOM.PL.FEM. that/*to-whom AUX

jim kdaj dali nagrado.

to-themDAT.PL.FEM. ever they-gave award

'Those are the best musicians that they have ever given (them) the award.'

Given the case assignment restrictions we have posited for Slovene in section 2.2.3, this

is entirely predictable. Because case cannot be phonetically realized on more than one D0

in a DP, dative case cannot be assigned to both the pronoun as well as a wh-relativizer.

Thus kar in these constructions cannot be a wh-relativizer.

Because kar in (130) and (131) behaves so differently from the wh-pronoun kar, it

is reasonable to treat the two as distinct (albeit polysemous) lexical entries. Given that

kar in (130) and (131) is morphologically invariable and its only function is to introduce

a clause, it is reasonable to conclude that, like ki, kar belongs to the lexical category of

complementizers rather than wh-pronouns.

In many ways, kar-phrases closely resemble ki-type relative clauses. Both

obligatorily prohibit relative wh-pronouns. Both obligatorily contain an indeclinable C0

as well as a pronoun whose interpretation is dependent on an antecedent and whose case

is determined by its function in the embedded clause. There are, however, two crucial

differences between the ki-type relative and the kar-phrase:

i. Kar-phrases modify only NPs whose heads are also modified by a

superlative degree adjective.75

75

There are examples of kar-phrases occurring without the presence of a superlative degree adjective, but

138

ii. The pronoun within the embedded kar-CP does not have complete

φ-feature agreement with the antecedent NP. Consider sentence

(132), in which the (presumed) NP associate, iskalec 'seeker', is

singular and the pronoun in the embedded kar-phrase, jih 'them' is

plural.

(132) Ti si najboljši iskalec, kar smo jih kdaj

you are the-bestNOM.SG.MASC. seekerNOM.SG.MASC. that AUX themACC.PL. ever

imeli!

we-had

'You are the best seeker that we have ever had (them)!' (Rowling 1999, trans. J.

Kenda)

These two defining properties of kar-phrases indicate a fundamental difference between

their syntax and the syntax of any of the relative clause types mentioned previously.

5.2.2. A Descriptive Overview of the Syntax of kar-phrases

The most salient property of kar-phrases is that they obligatorily include a

pronoun, which is almost always plural76

. The value of the number feature on the

modified N0 has no bearing at all on the value of the number feature on the pronoun in

rather with a comparative degree adjective taking scope over a universal quantifier (e.g. better than all) -

the semantics of this, as will be shown later, are identical to that of a superlative degree adjective. 76

There is one exception in which the pronoun contained in a kar-phrase appears in the singular, as in the

sentence below.

To je najslabše pivo, kar sem ga kdaj pila.

This is the-worstNOM.SG.NEUT. beerNOM.SG.NEUT. that AUX itACC.SG.NEUT. ever I-drank

'This is the worst beer that I ever drank (it).' (Chidambaram 2007)

This, as I will discuss later on in the chapter, is a function of the modified N0 being a mass noun.

139

the embedded CP. In example (133), the N0 is in the dual while the pronoun in the

embedded clause must be plural.

(133) Tisti sta najboljši studentki, kar (jaz)

thoseNOM.DU.FEM are the-bestNOM.DU.FEM. studentsNOM.DU.FEM. that I

sem jih / *ju učila.

AUX1st.SG. themACC.PL.FEM. themACC.DU.FEM. taughtSG.FEM.PAST.

'Those are the two best studentsdual that I taught (themplural /*dual).'

This fact about kar-phrases makes it impossible to derive them through the same syntax

that we have developed for relative clauses. Consider the following infelicitous derivation

of (133), modeled after the ki-type relative clause structure. Note that študentke is the

plural form of the word študentka (meaning studentNOM.SG.FEM.), whereas študentki is the

dual form.

140

Diagram 38

*TP

DP TP

Tisti

T VP

V DP

sta

D NP

Ø

AP NP

najboljši

N CP

študentkij

DPi CP

D NP C TP

OPWH študentkek kar

DP TP

(jaz)

T VP

sem

V DPi

učila

D DP

jihk

D NP

OPWH študentkek

In this derivation, the plural NP študentke 'studentsplural' merges with a silent wh-operator

D0. The pronoun then merges to this DP, thereby creating the stacked-DP. After the V

0,

T0, subject DP, and C

0 merge, the internal DP raises to Spec-CP, thereby stranding the

pronoun. The external relative head študentki 'studentsdual' is then merged. It is at this

point that the derivation runs into trouble. It is a condition on relative clauses that the

internal head must be co-referential with the external head; i.e., the external head must be

141

an antecedent of the internal head. In the structure presented in diagram (38), this

condition cannot be met: because the external head and internal head do not share the

same number feature (the external head is dual while the internal head is plural), they

cannot possibly be co-referential. This means that external relative head cannot be the

antecedent of the internal head, and therefore, Relative Deletion will not apply. The

resulting sentence would be:

(134) * Tisti sta najboljši študentki,

thoseNOM.DU.FEM are the-bestNOM.DU.FEM. studentsNOM.DU.FEM.

študentke kar (jaz) sem jih učila.

studentsACC.PL.FEM. that I AUX1st.SG. themACC.PL.FEM. taughtSG.FEM.PAST.

'Those are the two best students that I taught.'

Given the failure of this derivation to converge, the syntax of kar-phrases must be

different from that of ki-type relatives.

5.2.3. A Descriptive Overview of the Semantics of kar-phrases

The key to understanding the semantics of kar-phrases lies in their connection to

superlative degree adjectives. While the function of a relative clause is to identify and

describe an entity (be it an individual or a group), the function of a superlative clause (i.e.

a clause adjoined to an N0 modified by a superlative degree adjective) is to identify a

particular group from which the superlative degree adjective isolates a specific entity.

Consider again examples (126) and (127), repeated below.

142

(126) The seeker [that played for Hogwarts] caught the Golden Snitch.

(127) The best seeker [that played for Hogwarts] was Harry Potter.

In these two sentences, the embedded CP contains exactly the same words in exactly the

same order. Nevertheless, the two clauses differ in terms of their interpretations. For

example, only in (127) is a negative polarity item licensed in the embedded clause.

(126) a. * The seeker [ that ever played for Hogwarts] caught the Golden Snitch.

(127) a. The best seeker [ that ever played for Hogwarts] caught the Golden Snitch.

This means that the entailment environments of (126) and (127) are distinct; because the

superlative degree adjective quantifies over the left argument, (127) is left downward

entailing, whereas (126) is not.

Moreover, the embedded clause in (127) has a distinctly partitive reading.

Diagram 39

SEEKERS

HOGWARTS

QUIDDITCH

TEAM

143

Given sentence (126), the intersection of the two categories in diagram (39) would

consist of exactly one member. On the other hand, given sentence (127), the intersection

of the two categories in the diagram may be any number greater than 1. But in order for

the use of the superlative degree adjective not to be trivial, the intersection must be

greater than 177

. That is, if there were only ever one seeker on the Hogwarts quidditch

team, then to describe that seeker as "the best seeker that played for Hogwarts" is

tautological. That seeker will have been, by virtue of his uniqueness as a Hogwarts

seeker, both the best and the worst seeker of Hogwarts. Assuming, on the other hand, that

the intersection of the categories of seekers and Hogwart quidditch team members is >1,

it is the superlative degree adjective in (127) that identifies one particular member within

the intersection.

This distinction is significant because it demonstrates unequivocally that the

bracketed clause in (127) conveys a partitive meaning, whereas the superficially identical

clause in (126) does not. In other words, the bracketed clause in (127) could be

paraphrased as [among all the seekers that played for Hogwarts], whereas the same

bracketed phrase in (126) could not. The partitive kar-phrase in (127) in fact denotes a

comparison class from which the modified N0 head is singled out by the superlative

degree adjective (Chidambaram 2008). In order to determine how this is represented in

the syntax, it is helpful to look first at the derivation of a related construction.

77

As pointed out to me by Edwin Williams (p.c. 2013), the expectation that the intersection be >1 may only

be an implicature of the superlative rather than a genuine requirement, given the following example:

"The smallest positive integer less than n", when n = 2 describes only one possible value, namely 1.

However, if the statement were re-phrased not to include a variable, (i.e. "The smallest positive integer less

than 2"), the use of the superlative degree adjective would be tautological.

144

5.3. Comparatives in Slovene

There are two types of phrases that can adjoin to a comparative in Slovene: one is

a prepositional phrase headed by P od 'from', which checks accusative case on its object,

and the other is the Slovene analog to the English than-phrase, headed by C kot 'than'.

Diagram (40) represents the derivation of (135).

(135) Sašo je boljši študent od Mirkota.

SašoNOM.SG.MASC. is betterNOM.SG.MASC. studentNOM.SG.MASC. from MirkoGEN.SG.MASC.

'Sašo is a better student that Mirko.'

Diagram 40

TP

DP TP

Sašo

T VP

V DP

je

AP DP

boljši

DP compAP

študent

AP compAP

boljši

compA PP

-jš-

DP PP

AP DP P DP

boljši Sašo od Mirkota

145

In this derivation, the preposition od 'from' first merges with its object DP, Mirkota, where

it checks its genitive case. In the Specifier position of this PP is the DP boljši Sašo, which

contains the adjunct AP boljši 'better' and the DP Sašo. The adjunct AP checks its φ-

features with those of the DP. Both are masculine singular. The compA head merges to

PP. The compA head contains the features of the comparative affix78

; it is a phonetically

null head that has a [+comparative] feature and requires a PP complement headed by the

preposition od. The AP boljši raises into Spec-compAP in order to check its comparative

adjectival feature. The DP študent then merges and the AP boljši raises once more into

Spec-DP and checks its φ-features against študent. The copula merges followed by T0.

The DP Sašo then raises into Spec-TP to check its nominative case feature.

Alternatively, the compA head could take a CP headed by kot 'than' as its

complement.

(136) Sašo je boljši študent kot (je)

SašoNOM.SG.MASC. is betterNOM.SG.MASC. studentNOM.PL.MASC. than is

Mirko dober študent.

MirkoNOM.SG.MASC. goodNOM.SG.MASC. studentNOM.PL.MASC.

'Sašo is a better student than Mirko (is).'

78

I assume the Strong Lexicalist Hypothesis (Chomsky 1970): all words come into the derivation fully

formed but must check their lexical features during the course of the derivation. The affixal projection

compAP is not headed by the phonetic affix but rather the features of a comparative adjective that must be

checked by the adjective through a Spec-Head relation.

146

Diagram 41

TP

DP TP

Sašo

T VP

V DP

je

AP DP

boljši

DP compAP

študent

AP compAP

boljši

compA CP

-jš-

DP CP

AP DP C TP

boljši Sašo kot

DP TP

Mirko

T VP

V DP

je

DP DP

Mirko

dober študent

In this derivation, the DP Mirko first merges with an AP adjunct dober študent 'good

student'. Mirko raises past the copula to Spec-TP. The comparative C kot merges to form

CP. The DP boljši Sašo (containing the AP boljši 'better' and the DP Sašo) merges into

Spec-CP. From here, the derivation is identical to the previous derivation. The

comparative adjective raises to check it [+comp] feature in Spec-compAP then again to

147

check the φ-features of študent. The DP Sašo raises to Spec-TP of the matrix clause to

receive nominative case. At PF, however, the VP of the embedded clause is deleted by

Comparative Deletion (as originally proposed in Bresnan (1975) and elaborated by

Kennedy (2002) and Sauerland (2003) among others).

5.3.1. Comparatives and kar-phrases

While kar-phrases are most frequently found adjoined to N0 modified by a

superlative degree adjective, they can also occur with comparative adjectives, but only

when the comparison class includes a universal quantifier, as in (137):

(137) Ti si boljši iskalec od vseh

you are betterNOM.SG.MASC seekerNOM.SG.MASC from allGEN.PL.MASC

(iskalcev) kar (mi) smo jih imeli.

(seekersGEN.PL.MASC.) that we AUX1st.PL. themACC.PL. hadPL.PAST

'You are a better seeker than all that we have ever had (them).'

There is a striking semantic relationship between sentence (137), containing a

comparative and a universal quantifier, and sentence (131), repeated below as (138),

which contains a superlative. In fact, the two are synonymous.

(138) Ti si najboljši iskalec, kar smo jih kdaj

you are the-bestNOM.SG.MASC. seekerNOM.SG.MASC. that AUX themACC.PL. ever

imeli!

we-had

'You are the best seeker that we have ever had (them)!'

(Rowling 1999, trans. by J. Kenda)

148

The fact that these two syntactically distinct sentences have the same meaning indicates

that there exists an equivalency between their terms. That is, if we extract from these two

sentences the syntactic items that differentiate them (that is, items that occur in (137) but

not (138) and vice-versa), those items must be semantically equivalent. The relevant

phrases are:

(139) boljši iskalec od vseh

betterNOM.SG.MASC. seekerNOM.SG.MASC. of allGEN.PL.MASC.

'better seeker than all'

(140) najboljši iskalec

bestNOM.SG.MASC. seekerNOM.SG.MASC.

'best seeker'

(139) and (140) are semantically equivalent, which must mean that the following

equation is true:

boljši od vseh = najboljši

'better of all' = 'best'

In order to reduce this equation to its absolute simplest terms, it is useful first to consider

the morphology of comparatives and superlatives in Slovene.

149

5.3.2. The morphology of Slovene Comparatives and Superlatives

Slovene comparatives are typically constructed by the addition of the suffix -jš to

the adjectival root. The same -jš suffix is added even for adjectives that have suppletive

stems in the comparative.79

(141) pameten → pametnejši

'smartNOM.SG.MASC.' 'smarterNOM.SG.MASC.'

(142) dober → boljši

'goodNOM.SG.MASC.' 'betterNOM.SG.MASC.'

While comparatives are created by adding a suffix to the positive adjectival root,

the superlative is formed by adding the prefix naj- to the comparative adjectival stem.

(143) pameten → pametnejši → najpametnejši

smartNOM.SG.MASC. smarterNOM.SG.MASC. smartestNOM.SG.MASC.

(144) dober → boljši → najboljši

goodNOM.SG.MASC. betterNOM.SG.MASC. bestNOM.SG.MASC.

This is in line with Bobaljik's (2012) universal generalization of degree adjectival

morphology, which states that superlatives are universally a product of the morphological

Merger of the comparative stem with the superlative morpheme. Bobaljik notes

furthermore that even when the comparative form of the adjective is periphrastic (like

79

The final -i is the portmanteau ending for masculine nominative singular attributive adjectives.

Positive degree adjectives frequently have a zero ending for masculine singular nominative, particularly

when they are understood as indefinite, but all other degree adjectives must have the -i suffix.

150

English, some adjectives in Slovene cannot have a synthetic comparative form), the

superlative morpheme merges not with the adjectival root but rather with the comparative

adverb:

(145) zaspan → bolj zaspan → naj-bolj zaspan

sleepyNOM.SG.MASC. → more sleepyNOM.SG.MASC. → most sleepyNOM.SG.MASC.

If, following Bobaljik (2012), one concludes that the morphology of comparatives and

superlatives in Slovene is agglutinating, then the meaning attributed to each morpheme is

transparent. The -jš affix carries the meaning more and the naj- affix carries the meaning

of all. We can thus reduce the equation in section 5.3.1 to its most basic form:

od vseh = naj-

of all = -est

Intuitively, this seems to be a reasonable reduction, but moreover it is a logical equation

(see Chidambaram (2008) for a formal proof illustrating that the definition of a

comparative and the definition of a superlative differ only by the inclusion of a universal

quantifier in the latter). The link, then, between sentences like (137) and (138) is that they

both contain a universal quantifier; in (138), it is inherent in the superlative degree

morpheme, while in (137) it is phrased explicitly as a universal quantifier. Given that

these are the only conditions that license the kar-phrase, it is clear that kar-phrases occur

as a reflex of the presence of a universal quantifier80

. This claim is further substantiated

by the following data (Chidambaram 2007):

80

This is consistent with a suggestion made by Edwin Williams (p.c. 2013) that kar may be an NPI licenser,

151

(146) To so vsi iskalci kar smo

that are3rd.PL. allNOM.PL.MASC. seekerNOM.PL.MASC. that AUX1st.PL.

jih imeli

themACC.PL.MASC. hadPL.PAST

'Those are all the seekers that we have had (them).'

The occurrence of a kar-phrase is dependent upon presence of a universal quantifier.

5.3.3. A Formal Analysis of Comparatives and kar-phrases

The kar-phrase is similar in many ways to a relative clause: it is headed by a C0

and contains a clitic pronoun. In sentence (137) the antecedent of the clitic is the N0

iskalcev (seekersGEN.PL.MASC.). In the previous section, however, I showed that the kar-

phrase differs from a typical relative clause in that it can only occur as a complement of a

universal quantifier. Assuming, then, that the kar-phrase is a specialized partitive clause

sister to the universal quantifier, diagram (42) must be the derivation of sentence (137).

a possibility supported by the fact that kar-phrases can follow universal quantifiers such as every and each.

Laka (1990) presents evidence for the existence of other negative complementizers which license polarity

items. While I suspect that kar may indeed be a negative complementizer, as suggested by Williams, the

research needed to confirm the hypothesis extends beyond the reach of this dissertation.

152

Diagram 42

TP

DP TP

Ti

T VP

V DP

si

AP DP

boljši

DP compAP

iskalec

AP compAP

boljši

compA PP

-jš-

DP PP

AP DP P QP

boljši Ti od

Q DP

vseh

DP QP

(iskalcevi)

Q CP

vseh

DP CP

D NP C TP

OPWH iskalcevi kar

DP TP

(mi)

T VP

smo

V DP

imeli

D DP

jih

D DP

OPWH iskalcev

153

In this derivation, the NP iskalcev (seekersACC.PL.MASC.) merges with a silent wh-operator.

The clitic pronoun jih merges with this DP, creating a stacked-DP. The pronoun matches

its associate in reference and φ-features and checks its definiteness with its sister DP. The

verb, auxiliary, and the subject of the embedded clause then merge. After the C0 kar

merges to the TP, the internal DP (which contains the wh-operator and the pronominal

associate iskalcev) raises to Spec-CP to check the uninterpretable wh-feature on C. This

strands the pronoun jih. The N0 iskalcev (seekersGEN.PL.MASC.) followed by the universal

quantifier vseh then merge to the embedded CP. The DP iskalcev merges, then the Q0 vseh

raises to take scope over DP iskalcev. The preposition od 'from' merges to form the PP

and subsequently the subject DP and the comparative degree adjective merge into Spec-

PP. The comparative adjective affixal head then merges, and the comparative adjective

raises into Spec-compAP in order to check the affixal features. Then iskalec 'seeker'

merges and the comparative adjective raises once more to Spec-DP to check its φ-

features. The matrix verb then merges and the subject Ti (You) raises from Spec-PP to

Spec-TP. At PF the NP iskalcev within the kar-phrase may be deleted via an operation

similar to relative deletion; i.e., it is deleted under identity with the DP Spec-QP iskalcev.

This higher DP may also be elided at PF under near identity with the NP iskalec81

.

81

The occurrence of deletion on near-identity is common, although I do not provide an analysis of it here. It

does however occur in a variety of languages. Many thanks to Leonard Babby for the following example

from Russian:

Ja vzgljanul na te neskol'ko stranic prodšestvujušix vynutoj (stranice)

I looked at those few pagesGEN.PL.FEM. preceding ripped-outDAT.SG.FEM (pageDAT.SG.FEM)

im.

by-them

'I looked at those few pages preceding the one ripped out by them.'

154

This analysis explains the occurrences of a kar-phrase with overt universal

quantifiers. In sentence (138), however, there is neither an overt universal quantifier nor

an overt plural antecedent of the kar-phrase internal DP.

5.4. A Formal Analysis of Superlatives and kar-phrases

As shown above in section 5.3.2, the superlative morpheme itself contains the

universal quantifier, which makes it is possible for the kar-phrase to occur as the

complement of the superlative adjective affixal projection. Moreover, the semantics of

the superlative also offers insight into the obligatory plurality of the resumptive pronoun.

Because the kar-phrase is semantically a partitive CP denoting the domain of the

superlative degree quantifier, the requisite plurality on the pronoun is easily explained; as

noted in Chidambaram 2008, the plural DP acts as a partitive marker, and therefore, it

must be plural (deHoop 2003).

Having derived the syntax of Slovene comparatives, the relationship between

comparatives and superlatives, and the restrictions on kar-phrases, we are now in a

position to propose an explicit syntactic analysis of (138), repeated below.

(138) Ti si najboljši iskalec, kar smo jih kdaj

you are the-bestNOM.SG.MASC. seekerNOM.SG.MASC. that AUX themACC.PL. ever

imeli!

we-had

'You are the best seeker that we have ever had (them)!'

(Rowling 1999, trans. by J. Kenda)

In this example, the singular form of the word page (stranice) is deleted on near identity with a higher

lexical item, namely the plural genitive form pages (stranic).

155

Diagram 43

TP

DP TP

Ti

T VP

V DP

si

AP DP

najboljši

DP compAP

iskalec

AP compAP

najboljši

compA supAP

-jš-

AP supAP

najboljši

supA CP

naj-

DP CP

D NP C TP

OPWH iskalcev kar

DP TP

(mi)

T VP

smo

V DP

imeli

D DP

jih

D NP

OPWH iskalcev

156

This derivation has a great deal in common with the structure of (137) as illustrated in

diagram (42): The embedded kar-phrase CP here is identical to the one in (137). The

chief difference between the two is that the head that merges to the maximal kar-CP in

(137) is the universal quantifier vseh (all) whereas in (138), it is the superlative affixal

head that takes the kar-CP as its complement82

. The AP najboljši 'best' then merges as the

Specifier of the superlativeAP (i.e. supAP), where it checks its superlative features

against the superlative affixal head, naj-. The comparative affix then merges and the AP

raises to Specifier of compAP to check its comparative features against the comparative

affixal head, -jš-. The DP iskalec then merges and the adjective najboljši raises once more

to Spec-DP in order to check its φ-features. Finally, the matrix verb and subject merge

into the derivation. At PF, the DP iskalcev is obligatorily deleted under near identity with

the higher DP iskalec. This deletion operation could be due to a Distinctness Condition

(as proposed by Richards 2000), which does not allow two items of the same label to be

in linearly adjacent at PF.

5.5. The Resumptive Pronoun in kar-phrases

Kar-phrases are distinct from relative clauses, since they are not modifiers of an

N0 but rather they are partitive CP complements of a universal quantifier, but they

nevertheless have have some characteristics in common with relative clauses. Both

82

The notion that the superlative adjective bears a special relationship to the CP is also supported by Cinque

(2010) who shows that a superlative degree adjective always takes scope over the CP, while a modal

adjective may not. Compare the following sentences:

a. Harry is the youngest Gryffindor that will enter the Tri-Wizard Tournament.

b. Harry is the probable Gryffindor that will enter the Tri-Wizard Tournament.

While youngest necessarily takes scope over the CP (Gryffindor) that will enter the Tri-Wizard Tournament,

the same is not true for the modal adjective probable, which only takes scope over Gryffindor. Many thanks

to Edwin Williams for pointing out these supporting data.

157

contain a gap in their English translations (see sentences (126) and (127)), although the

gap is not interpreted in the same way for both. I have suggested that both obligatorily

contain a pronoun and a wh-operator. In the kar-CP, the C0 must be pronounced (i.e. kar

is unrecoverable and thus cannot be deleted), therefore, the wh-operator is invariably

silent. The operator checks the definiteness on the pronoun and then can raise into Spec-

CP, thereby stranding the pronoun.

5.6. The Interaction with Mass Nouns

We have seen above that the pronoun in a kar-phrase is always marked as plural.

The one apparent exception occurs when the associate N0 is a mass noun. Only then can

the pronoun occur in the grammatical singular:

(147) Najslabše pivo je, kar sem ga kdaj

worstNOM.NEUT.SG. beerNOM.NEUT.SG. is that AUX1st.SG. itACC.NEUT.SG. ever

pila.

drankFEM.SG. PAST

'That is the worst beer that I ever drank (it).' (Chidambaram 2007)

Given the analysis presented so far, there is nothing surprising or inexplicable about this.

The superlative prefix naj- licenses the kar-phrase, which is semantically partitive. But

because the quantified noun is a mass noun, the result is that the resumptive pronominal

paritive marker occurs in the grammatical singular. The associate of the resumptive

pronoun is also the grammatical singular pivo 'beer', however it is semantically plural;

i.e., in the CP kar sem ga kdaj pila, the understanding is that the speaker has in the past

consumed many cans/bottles/glasses of beer and the resumptive pronoun ga refers to all

158

the beer she previously consumed. Of all the beer the speaker has consumed, this

particular can/glass/bottle of beer is the worst.

If, however, the speaker wanted to indicate that this particular variety/brand of

beer was the worst that she ever drank, the sentence would be:

(148) Najslabše pivo je, kar sem jih kdaj

worstNOM.NEUT.SG. beerNOM.NEUT.SG. is that AUX1st.SG. themACC.PL. ever

pila.

drankFEM.SG. PAST

'That is the worst (kind of) beer that I ever drank (them).'

(Chidambaram 2007)

As noted in Chidambaram (2007), this "variety/brand" reading is typical of a pluralized

mass noun. When someone mentions beers, they often mean types or brands of beer, as

opposed to the singular beer, which could refer to any quantity of beer and the

variety/brand is irrelevant. In (129), the associate of the resumptive pronoun would be

pivaACC.PL.NEUT. 'beers'.

5.7. Languages other than Slovene

Slovene appears to be unique among the world's languages in having an overtly

plural resumptive pronoun in clauses selected by a universal quantifier. Many other

languages, both related and unrelated to Slovene, employ resumption in relative clauses,

but none of them has the option of including an overt plural resumptive pronoun in their

equivalents of kar-phrases. Interestingly, however, it appears that in many languages that

exhibit canonical resumption, the equivalent of kar-CPs are precluded from containing

any sort of resumptive pronoun at all.

159

As shown in th previous chapter, Slovak has canonical resumption in relative

clauses. Consider the following set of Slovak sentences:

(149) Stíhač, čo ho máme, sa volá Harry

seekerNOM.SG.MASC. that himACC.SG.MASC. we-have REFL call3rd.SG.PRES. Harry

Potter.

Potter

'The seeker that we have (him) is named Harry Potter.'

(150) Si najlepší stíhač, čo sme *ho

you-are the-bestNOM.SG.MASC. seekerNOM.SG.MASC. that AUX1st.PL. *himACC.PL

/*ich kedy mali.

/*themACC.PL. ever we-had

'You are the best seeker that we have ever had *him/*them.'

(151) Si najlepší stíhač, čo sme kedy mali.

you-are the-bestNOM.SG.MASC. seekerNOM.SG.MASC. that AUX1st.PL. ever we-had

'You are the best seeker that we have ever had.'

Sentence (149) contains a relative clause with a resumptive pronoun, which as expected,

agrees with its overt antecedent in all φ-features as well as its reference. In other words,

ho in (149) behaves like a typical resumptive pronoun. Resumptive pronouns in Slovak

must fully match the antecedent's feature complex, otherwise the sentence is

ungrammatical.

(150) contains the Slovak equivalent of a Slovene superlative followed by a kar-

phrase. Unlike Slovene, however, Slovak does not license the presence of any pronoun as

a resumptive within the clause subordinate to the superlative. Neither a singular pronoun

160

matching the φ-features and reference of the NP in the higher clause nor a plural pronoun

can appear in the embedded CP. The only option is (151), in which there is no resumptive

pronoun present.

The inability of the embedded clause in (150) to support resumption offers a few

insights and invites many questions. First of all, it reveals that while the embedded CP in

(150) and (151) appears similar to the relative clause in (149), the two clauses must be

distinct. That is, the embedded clause in (150) and (151) cannot be an ordinary relative

clause; if it were, it would license a resumptive pronoun. It is crucial to note that (150)

and (1151) include a superlative. This leads to the following hypothesis: perhaps Slovene

and Slovak are structurally similar in terms of their superlative licensed CPs but differ

only in that Slovene allows a resumptive pronoun to be pronounced when its associate is

elided at PF and Slovak does not.

The pronunciation of resumptive pronouns in Slovene is obligatory. In ki-phrases

in which the resumptive functions as anything but the highest subject, the resumptive

pronoun must be pronounced. But in many other languages that exhibit resumption, such

as Slovak, the use of resumptives is optional under certain conditions. It follows that

Slovene has different requirements for the pronunciation of resumptives than Slovak. It is

not entirely surprising, then, that the resumptive pronoun in the Slovene kar-phrase is

obligatorily pronounced while in the equivalent Slovak phrase it cannot be.

The possibility that, in terms of these kar-phrases, Slovak is structurally identical

to Slovene in all but its PF, gives rise to the hypothesis that all languages are like

Slovene, in spite of their surface differences. If we look again at the English sentences in

161

(126) and (127), repeated here as (152) and (153), there are noticeable differences in the

bracketed clauses, although the PF-string seems, at first, to be the same.

(152) The seeker [that played for Hogwarts] caught the Golden Snitch.

(153) The best seeker [that played for Hogwarts] was Harry Potter.

In section 5.2.3, I showed that these two clauses are semantically distinct, but it seems

that there is a significant PF-distinction as well: Their intonational patterns are different83

.

In (152) the intonation of the word Hogwarts is high, low-rise. The first syllable is

heavily accented, i.e. the pitch and volume are high. The second syllable is pronounced

low at the onset and the pitch rises near the coda. In (153) the word Hogwarts is

unstressed. The first syllable has the same pitch and volume as the rest of the embedded

clause. Near the coda of the second syllable, the pitch is slightly raised. Put simply, there

is contrastive stress on Hogwarts in (152) whereas there is no such contrastive stress on

the same word in (153)84

.

The evidence from both the semantics of these phrases and their distinct

intonational patterns suggests that in English, as in Slovene and Slovak, the two are not

structurally identical. While (152) contains an ordinary relative clause, (153) contains the

83

I ran an short experiment asking speakers to read these sentences among several other sentences as I

recorded them. All six speakers come from the mid-Atlantic region of the United States, and thus are from

the same dialect group, which may have some effect on the data. All six speakers gave the same results.

The experiment, whose specifics are detailed in Appendix A, was informal and thus its results may be

flawed, but given the unanimaty of the findings, it is worth mentioning the results and later conducting a

more formal experiment. 84

As pointed out by E. Wayles Browne, contrastive stress can be placed on the word Hogwarts even in the

English equivalent of the kar-phrase, however the contrast must be set up in the discourse - it is not

inherently part of the sentence. The inherent contrast being made in the kar-phrase is the contrast between

this particular seeker and all other seekers at Hogwarts, not between seekers at Hogwarts and seekers at a

different school.

162

English equivalent of a kar-phrase. English, unlike Slovak, does not have resumptive

pronouns at PF under any conditions85

, therefore no resumption occurs in the English

equivalent of a kar-phrase.

5.8. ki-relatives with superlative modified DPs

Another apparent problem with the analysis of kar-phrases presented here is that

Slovene has the option of allowing a DP modified by a superlative to take a ki-type

relative clause. As mentioned in section 5.1.3, the ki-type relative obligatorily includes a

resumptive pronoun in any position where resumptives are available (except the highest

subject position), but that the resumptive must match its antecedent in all its φ-features.

In the previous sections, I claimed that kar-phrases are, in fact, no different from ki-

relatives in this regard but differ from them only insofar as their antecedents are

unpronounced; while the antecedent of the resumptive in a ki-type relative is obligatorily

overt, the antecedent of a resumptive in a kar-phrase is obligatorily unpronounced. That

aside, the two clause types have a similar structure; the primary distinction is that the kar-

phrase is an argument of supA, while a ki-relative is an adjunct of N.

The inevitable question, then, is why sentences like (154) are admissible in

Slovene:

(154) Ti si najboljši iskalec, ki smo ga

you are bestNOM.SG.MASC. seekerNOM.SG.MASC. that AUX1st.SG.PRES. himACC.SG.MASC.

imeli.

had

85

As mentioned in chapter 1, footnote 2, some speakers of English do employ intrusive pronouns, but these

are categorically distinct from resumptives.

163

'You are the best seeker that we have (him).'

If kar-phrases are an argument of supA, a syntactic head which is clearly present in (154),

and they obligatorily contain a plural resumptive due to the internal structure and

semantics of the kar-phrase, what are the structure and semantics of the ki-relative in

(154)? The solution is suggested by the fact that (155) is also grammatical.

(155) Ti si najboljši iskalec.

you are bestNOM.SG.MASC. seekerNOM.SG.MASC.

'You are the best seeker.'

Given the grammaticality (155), it is clear that the kar-phrase is PF-optional: Although at

LF there must be a domain over which the superlative quantifies (i.e., a kar-phrase must

be present at the level of interpretation), there is no need for it to be pronounced.

Thus the answer to the question of what happens when we find a ki-relative rather

than a kar-phrase paired with a superlative is that the ki-relative does not replace the kar-

phrase either in form (as a sister of supA) or in function (as a descriptor of the quantified

domain) but rather may occur alongside it, as shown in diagram (44). The ki-relative in

(154) is, then, no different from any ordinary ki-relative and has no special relationship to

the superlative degree adjective, whereas the kar-phrase has a close relationship (head-

complement) with supA. I conclude, therefore, that the derivation of (154) is significantly

different from the derivation of (138), in that the ki-type relative clause is a canonical

relative clause: it is a CP sister to N.

164

Diagram 44

TP

DP T'

Ti

T VP

V DP

si

AP DP

najboljši

DP compA...

D NP

Ø N CP

iskalec

DP C'

OPWH iskalec

C TP

ki

DP T'

(mi)

T VP

smo

V DP

imeli

DP DP

OPWH iskalec

D DP

ga OPWH iskalec

165

The rest of the derivation (indicated by the ellipsis next to compA) is no different from

the one in diagram (43). Slovene allows both the ki-relative as well as the kar-phrase to

be pronounced in a single sentence and they can be pronounced in either order86

.

(156) Ti si najboljši iskalec, ki sem ga osebno

you are the-best seeker that AUX1st.SG.PRES. himACC.SG.MASC. personally

spoznala, kar smo jih imeli na

metSG.PAST.FEM. that AUX1ST.PL.PRES. themACC.PL.MASC. hadPL.PAST. at

Bradavičarki!

Hogwarts

'You are the best seeker that I have met (him) in person that we have had (them) at

Hogwarts!'

(157) Ti si najboljši iskalec, kar smo jih imeli na

you are the-best seeker that AUX1ST.PL.PRES. themACC.PL.MASC. hadPL.PAST. at

Bradavičarki, ki sem ga osebno spoznala!

Hogwarts that AUX1st.SG.PRES. himACC.SG.MASC. personally metSG.PAST.FEM.

'You are the best seeker that we have had (them) at Hogwarts that I have met (him)

in person!'

The two syntactic constructions (superlative+kar-phrase and superlative+ki-relative) have

slightly different interpretations, which has to do with contrastive focus: the partitive

nature of the kar-phrase (i.e., the fact that it inherently contains a comparison domain),

makes the kar-phrase fundamentally contrastive, whereas the ki-relative does not convey

strong contrastive focus.

86

The order shown in (157) is actually preferable to the order shown in (156), even though the order in

(156) is the linear order derived in narrow syntax. The reason for this is possibly linked to an issue of

information structure; that is, the information is easier to parse when presented in the order shown in (157).

166

5.9. Summary

The clearest and least controversial conclusion drawn in this chapter is that

Slovene kar-phrases are not the same as ordinary Slovene relative clauses. Given the data

alone and without positing any more precise or abstract differences, the claim that the two

CPs are different from one another is undeniable. While there are certain similarities

between the two constructions, namely, they are both CPs in which wh-movement of

some DP to Spec-CP is obligatory, the differences between them are significant.

The primary purpose of this chapter was to investigate the properties of the

Slovene kar-phrase, which is a phrase containing a pronominal element similar to but

distinct from a canonical resumptive pronoun, since the pronoun in a kar-phrase does not

necessarily agree with its overt associate in its number feature. Three natural questions

arise: 1) Why is the pronoun in the kar-phrase invariably plural?, 2) How does the

plurality on the kar-phrase pronoun arise?, and 3) Is the kar-phrase pronoun a resumptive

pronoun? In this chapter, I have answered all of these questions.

In order to answer the first question, we turn to the semantics of the phrase. We

find that the semantics of the kar-phrase are distinct from those of an ordinary relative

clause, and that the kar-phrase denotes the class of entities over which the universal

quantifier takes scope. In some cases, the universal quantifier is overt, but in many cases,

as in superlative degree adjectives, the universal quantifier surfaces as the superlative

degree morpheme. It is either the superlative degree morpheme or the universal quantifier

that merges with the kar-phrase. Because the kar-phrase is a partitive CP, the pronoun in

it is invariably plural since it is the partitive marker.

167

The answer to the second question - how the plurality of the kar-phrase pronoun

arises - must lie in the syntax of the phrase. The structure of the kar-phrase is similar to

that of an ordinary relative clause. The crucial distinction between them is that in a kar-

CP, the equivalent of the internal relative head DP is not pronounced. The kar-phrase is a

CP whose head is the invariable/indeclinable C kar and which contains an obligatorily

plural DP object. The structure of this DP object is the same as the structure which gives

rise to canonical resumptive pronouns in ordinary relative clauses; it has the stacked-DP

structure including a pronoun, a wh-operator, and the pronominal associate. The pronoun

is merged as a determiner to the plural DP. The internal DP subsequently raises to Spec-

CP of the kar-phrase. Following an operation similar to relative deletion, Spec-CP of the

kar-phrase is obligatorily deleted under near identity with a higher DP. The reasons

behind the PF-deletion of the DP Spec-CP are not fully clear; this issue requires further

investiagtion.

Nevertheless, it seems that the answer to the third question follows naturally from

the answers to the first two questions: the plural pronoun in the kar-phrase is indeed a

resumptive pronoun whose associate is unpronounced. The kar-phrase internal pronoun

occurs in the same syntactic environment as a canonical resumptive pronoun; i.e., it

merges as a D to its DP associate and must satisfy the Pronominal Reference Condition.

The associate raises, thus stranding the pronoun (Boeckx 2003). Given the analysis

presented in this chapter, the obligatorily plural pronoun that occurs in the kar-phrase is a

resumptive pronoun.

Finally, I submit that the Slovene kar-phrase is not unique. While its structure

does not surface transparently in other languages, both the semantic and phonetic

168

characteristics of corresponding phrases in other languages suggest that they too are

structurally closer to the Slovene kar-phrase than to an ordinary relative clause. In this

chapter, I have been able to show that, at the very least, these clauses are distinct from

ordinary relaive clauses in several languages besides Slovene. In order to prove or

disprove this hypothesis (that is, that the structure of all kar-phrases cross linguistically is

the same), more research must be conducted into the various manifestations of the kar-

phrase in other languages.

169

CONCLUSION

This thesis is a step towards a generalized theory of pronouns. I have developed a

theory of pronouns that accounts for the basic binding properties of ordinary pronouns as

well as many of the properties of resumptive pronouns. The stacked-DP analysis of

pronouns, which combines elements of Boeckx's (2003) analysis of resumptive pronouns

and Franks and Rudin's (2005) analysis of clitic doubles, applies to all pronouns. This

eliminates one of the chronic problems in previous studies of resumptive pronouns,

namely, that resumptive pronouns are treated as lexically and/or syntactically distinct

from ordinary pronouns.

The stacked-DP analysis of pronouns is largely based on Freidin and Vergnaud's

(2001) analysis of definite pronouns as definite descriptions. From their original concept,

I develop a syntactic structure of the pronoun, which is a complex DP, containing the

pronoun as a D0 and its definite DP sister, which in turn contains a definite D

0 and the

pronoun's NP referent.

I present a new strategy for deriving canonical resumptive pronouns that

combines the stacked-DP analysis of pronouns with Sauerland's matching model of

relative clauses. This analysis differs from all previous analyses of resumptive pronouns

and relative clauses in that it introduces the notion that all relative clauses cross-

linguistically contain a pronominal element merged at the site of relativization. I advance

the possibility that the only difference between languages that have resumptive pronouns

and those that do not is that in the latter, the internal DP is of a stacked-DP is inert, i.e., it

cannot raise on its own.

170

I propose two new syntactic rules: a PF deletion rule and an LF interpretation

rule. The PF rule is Pronominal Associate Deletion, which deletes the associate of the

pronoun at PF when it is sister to the pronominal D0 head. If the associate raises prior to

Spell-Out, the conditions for Pronominal Associate Deletion will not be met. This is,

ultimately, the source of all resumption. The LF rule I propose is the Pronominal

Reference Condition, which forces the co-reference of the pronoun and its associate.

Because this rule applies at LF, it is irrelevant whether the associate undergoes movement

in the narrow syntax, because the reference can be derived via the relationship of the

pronoun and a copy of the associate. In addition to these two rules, I suggest that Relative

Deletion is a narrow syntactic operation that deletes any element in the Specifier of a

relative CP which takes the external relative head as its antecedent.

These three syntactic requirements (i.e. Relative Deletion, Pronominal Associate

Deletion, and the Pronominal Reference Condition) along with the stacked-DP analysis

of pronouns and the Sauerland matching model of relative clauses will generate structures

containing canonical resumptive pronouns. Furthermore, I show that the three syntactic

requirements combined with the stacked-DP analysis can also derive Macedonian clitic-

doubling as well as resumptive pronouns at the site of interrogative wh-movement.

Finally, I also show how this analysis accounts for certain idiosyncratic phenomena,

specifically optional deletion of canonical resumptive pronouns in B/C/S and Slovak and

resumptive pronouns in superlative clauses (kar-phrases) in Slovene.

While the analysis proposed in this thesis accounts for much of the data related to

resumption in the Slavic languages, further research is required to show how it accounts

for other instances of resumption and how well the stacked-DP analysis can account for

171

other phenomena related to pronouns (i.e. binding phenomena). The proposal presented

here has broad implications: it offers a general structural analysis applying to all definite

pronouns and furthermore reduces several phenomena (canonical resumption, clitic-

doubling, wh-interrogative resumption, superlative clause resumption) to the same basic

operations. Precisely because I have attempted to produce a broad analysis, the

consequences of my analysis are also necessarily broad; a great deal more research is

needed to show how this analysis accounts for all the extensive data to which it is

relevant.

172

APPENDIX A

Experiment Title: The Prosodic Features of the English relative clause vs. the kar-

phrase87

Objective: To determine whether there is a difference between the phonetic forms of an

English relative clause and the English equivalent of a kar-phrase88

.

Equipment: Olympus Voice Recorder

Design: The participant receives a sheet of paper with 7 sentences. Each sentence is

numbered. There are several lines of space between the sentences. The participant is

instructed to read each sentence to himself/herself prior to reading the sentence aloud and

to wait for two seconds between sentences. The relevant sentences are interspersed with

sentences that have no bearing on the experiment so that the participant will not guess at

what is being tested. The sentences are randomly ordered.

Test Subjects: There are six test subjects: three males and three females. All of them are

monolingual native speakers of American English. Two of the subjects (1 male and 1

female) are left-handed. All six live in the Mid-Atlantic region of the United States. They

range between the ages of 33 and 70.

87

This experiment was performed in full compliance with all guidelines required by the Office of Human

Subjects Research Protection. 88

A kar-phrase is defined as a CP adjoined to a superlative modified NP. It is a CP possessing the distinctive

semantic characteristics of a kar-phrase as outlined in section 5.2.3.

173

Test Material: The sentences they were given to read are listed below. The ones that are

relevant to the experiment are marked with an asterisk. The sentences were separated

from one another by several lines and had no punctuation besides a period at the end of

each sentence.

*1) John is the student that I met yesterday.

2) The book that John gave me was Harry Potter.

3) The tri-wizard tournament begins tomorrow.

*4) The seeker that played for Hogwarts caught the Golden Snitch.

5) The seeker for Gryffindor is Harry Potter.

*6) John is the smartest student that I met at the conference.

*7) The best seeker that played for Hogwarts was Harry Potter.

Results: The clearest results came from comparing the intonation of sentences (4) and

(7). For these sentences, and specifically for the intonation on the CP that played for

Hogwarts, all of the participants followed a similar intonational pattern.

For the CP in sentence (4), the pitch and volume of the word Hogwarts,

particularly on the first syllable of the word Hogwarts, is raised. The second syllable has

low rising pitch. The intonation of the same string of words in (7) has noticeably different

intonation. The first sylable of Hogwarts is spoken at the same pitch and volume as the

final syllable of the preceding word - there is no additional stress. The second syllable has

low-rising pitch.

174

Conclusions: The term Hogwarts in sentence (4) is marked by contrastive stress, while

the same term in (7) is not. Thus, the two CPs that appear identical (i.e. that played for

Hogwarts) are, in fact, distinct both at PF and LF. This suggests that the two actually

differ from one another within the narrow syntax; that is they are structurally distinct

CPs.

Chidambaram 175

Bibliography

Andrews, Avery D. 1985. Studies in the Syntax of Relative and Comparative Clauses.

New York: Garland.

Aoun, Joseph, Lina Choueiri, and Norbert Hornstein. 2001. "Resumption, Movement,

and Derivational Economy". Linguistic Inquiry 32(3): 371–403.

Asudeh, Ash. 2011. "Towards a unified theory of resumption". In Alain Rouveret, ed.,

Resumptive Pronouns at the Interfaces. Amsterdam: John Benjamins. 121–187.

Babby, Leonard H. 2009. The Syntax of Argument Structure. Cambridge, UK: Cambridge

University Press.

Babby, Leonard H. 1987. "Case, Prequantifiers, and Doscontinuous Agreement in

Russian." Natural Language and Linguistic Theory 5, 91-138.

Babby, Leonard H. 1986. "The Locus of Case Assignment and the Direction of

Percolation: Case Theory in Russian". In Brecht, Richard and James Lavine, eds.,

Case in Slavic. Columbus, OH: Slavica. 170-219.

Bhatt, Rajesh. 2002. "The raising analysis of relative clauses". Natural Language

Semantics, 10.

Bianchi, Valentina. 2004. "Resumptive Relatives and LF Chains". The Structure of CP

and IP: The Cartography of Syntactic Structures, Vol. 2., Luigi Rizzi, ed. Oxford:

Oxford University Press.

Bianchi, Valentina. 2000. "The Syntax of Relative Determiners". in The Syntax of

Relative Clauses. A. Alexiadou, P. Law, A. Meinunger, C. Wilder, eds. Amsterdam

/ Philadelphia: John Benjamins Publishing Company.

Bobaljik, Jonathon. 2012. Universals in Comparative Morphology. Cambridge, MA: the

MIT Press.

Boeckx, Cedric. 2003. Islands and Chains: Resumption as Stranding. Amsterdam: John

Benjamins Publishing Company.

Borer, Hagit. 1984. "Restrictive Relatives in Modern Hebrew". Natural Language and

Linguistic Theory. Vol. 2: 219-260. the Netherlands: Kluwer Academic

Publishers.

Bošković, Željko. 2009. "On Relativization Strategies and Resumptive Pronouns".

Proceedings of Formal Description of Slavic Languages 7, 79-92.

Chidambaram 176

Bresnan, Joan. 1973. "Syntax of the Comparative Clause Construction in English".

Linguistic Inquiry 4: 275-343.

Bresnan, Joan. 1975. "Comparative Deletion and Constraints on Transformations".

Linguistic Analysis 1: 25-74.

Browne, E. Wayles. 1986. "Relative Clauses in Serbo-Croatian in Comparison with

English". New Studies Series, Vol. 4. Ed. Filipović, Rudolf. Institute of

Linguistics, University of Zagreb. Zagreb. Yugoslavia.

Cardinaletti, A. and M. Starke. 1999. "The typology of structural deficiency: a case study

of the three classes of pronouns". Clitics in the Languages of Europe. 145-245. H.

Van Reimsdijk, ed. New York: Mouton de Gruyter.

Cecchetto, C. 2000. "Doubling structures and reconstruction". Probus 12:93-126.

Chidambaram, Vrinda. 2008. "The Subcategorization and Quantification Properties of

Superlatives in Slovene". Proceedings of Formal Approaches to Slavic Linguistics

16: the Stony Brook Meeting, 84-101 Bailyn, J., Bethin, C., and Antonenko, A.,

eds. . Ann Arbor, MI.: Michigan Slavic Publications.

Chidambaram, Vrinda. 2007. "Relative and Pseudo-RElative Clauses in Slovene".

Slovene Linguistics Studies 6: 287-301. Greenberg, M. and Snoj, M., eds. Inštitut

za slovenski jezik Frana Ramovša. Ljubljana, Slovenia.

Cinque, Guglielmo. 2010. The Syntax of Adjectives: A Comparative Study. Cambridge,

MA: the MIT Press.

Chomsky, Noam. 2008. "On Phases". In Foundational Issues in Linguistic Theory:

Essays in Honor of Jean-Roger Vergnaud. Cambridge, MA: the MIT Press.

Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: the MIT Press.

Chomsky, N. and Lasnik, H. 1993. "The theory of principles and parameters". Syntax: An

International Handbook of Contemporary Research, 506-569. Jacobs, J., von

Stechow, A., Sternefeld, W., and Vennemann, T., eds. Berlin: Mouton de Gruyter.

Chomsky, N. 1993. "A minimalist program for linguistic theory". In The view from

Building 20, K. Hale and S.J. Keyser (eds), 1-52. Cambridge, MA.: MIT Press.

[Re printed in Chomsky, N. 1995 The Minimalist Program, 167-217.]

Chomsky, N. and Lasnik, H. 1977. "Filters and Control". Linguistic Inquiry 8: 425-504.

Chomsky, N. 1970. "Remarks on Nominalizations." Readings in English

Transformational Grammar, 184-221. Jacobs, R. and Rosenbaum, P., eds.

Waltham MA: Ginn & Co.

Chidambaram 177

Clemens, L., Morgan A., Polinsky M., and Xiang M. 2012. "Listening to Resumptives:

An Auditory Experiment", 14-17 March. CUNY Sentence Processing Conference.

New York, NY: CUNY.

de Hoop, Helen. 2003. "Partitivity". The Second Glot International State-of-the-Article

Book: The Latest in Linguistics, 179-212. Cheng, L. and Sybesma, R. Berlin:

Mouton de Gruyter.

Dimitrova-Vulchanova, Mila and Giusti, Giuliana. 1998. "Fragmants of Balkan Nominal

Structure". Possessors, Predicates, and Movement in the Determiner Phrase.

Artemis Alexiadou and Chris Wilder, eds. Linguistik Aktuell / Linguistics Today.

Doron, Edit. "On the Syntax and Semantics of Resumptive Pronouns. 1982. Texas

Linguistics Forum 19: 1-48. University of Texas, Austin.

Erteschik-Shir, Nomi. "Resumptive Pronouns in Islands". 1992. Island Constraints: 89-

108. H Goodluck and M. Rochemont, eds. Dordrecht: Kluwer Academic

Publishers.

Fiengo, R., and May, R.. 1994. Indices and Identity. Cambridge, MA: MIT Press.

Franks, Steven. 2009. "Macedonian Pronominal Clitics as Object Agreement Markers". A

Linguist's Linguist: Studies in South Slavic Linguistics in Honor of E. Wayles

Browne, 189-221. Franks, S., Chidambaram, V., and Joseph, B. eds.

Bloomington,IN: Slavica Publishers.

Franks, Steven. 2008 "Clitic Placement, Prosody, and the Bulgarian verbal complex"

Journal of Slavic Linguistics 16 (1): 91-137.

Franks, Steven. 2006. "Another look at li placement in Bulgarian". The Linguistic Review

23:161–. 211

Franks, S. and Rudin, C. 2005. "Bulgarian Clitics as K0 Heads" The Proceedings of

Formal Approaches to Slavic Linguistics 13: the South Carolina Meeting, 104-

166. Franks, S., Gladney F.Y., and Tasseva-Kurktchieva, M.P., eds. Ann Arbor,

MI: Michigan Slavic Publications.

Franks, Steven. 1999. "Optimality Theory and clitics at PF". Formal Approaches to

Slavic Linguistics: The Seattle Meeting 1998, 101-116. Dziwirek, K., Coats, H.,

and Vakerliyska, C. eds. Ann Arbor: Michigan Slavic Publications.

Franks, Steven. 1998. "Clitics in Slavic". Paper presented at the Comparative Slavic

Morphosyntax Workshop. Bloomington, IN: Indiana University.

Chidambaram 178

Franks, Steven. 1995. Parameters of Slavic Morphosyntax. Oxford, UK: Oxford

University Press.

Freidin, Robert, and Jean-Roger Vergnaud. 2001. "Exquisite connections: Some remarks

on the evolution of linguistic theory". Lingua 111:639-666.

Freidin, Robert. 1997. Generative Grammar: Theory and its History. London, U.K.:

Routledge Leading Linguistic Series.

Gierling, Diana. 1997. "Clitic Doubling, Specificity, and Focus in Romanian". Clitics,

Pronouns, and Movement. eds. Black, James and Virginia Motapanyane.

Philadelphia, PA and Amsterdam, the Netherlands: John Benjamins.

Goodluck, Helen and Stojanovic, Danijela. 1996. "The Structure and Acquisition of

Relative Clauses in Serbo-Croation". Language Acquisition 5, 285–315.

Gračanin-Yuksek, Martina. 2010. "On a Matching Effect in Headed Relative Clauses".

Proceedings of Formal Approaches to Slavic Linguistics 18: the Cornell Meeting,

193-209. Browne, W., Cooper, A., Fisher, A., Kesici, E., Predolac, N., and Zec,

D., eds. Ann Arbor, MI: Michigan Slavic Publications.

Greenberg, Marc. 1999. "Multiple Causation in the Spread and Reversal of a Sound

Change: Rho-tacism in South Slavic". Slovenski Jezik / Slovene Linguistic

Studies, 63-76. Ljubljana / Lawrence.

Grohmann, Kleanthes and Panagiotidis, Phoevos. 2004. "Demonstrative Doubling in

Greek". University of Maryland Working Papers in Linguistics 13, 109-131.

College Park, MD.

Groos, A. and H. van Riemsdijk. 1979. "Matching Effects in Free Relatives: A Parameter

of Core Grammar". Theory of Markedness in a Generative Grammar:

Proceedings of the IV GLOW Conference. A. Belletti et al, eds. Pisa.

Halpern, A. 1992. Topics in the Syntax and Placement of Clitics. Doctoral Dissertation,

Stanford University.

Heim, Irene. 1987. "Where does the Definiteness Restriction Apply? Evidence from teh

Definiteness of Variables." In The Representation of (In)definiteness, 21-42. ter

Meulen, A. and E. Reuland, eds. Cambridge, MA: MIT Press.

Hladnik, Marko. 2010. Restrictive Relative Clauses in Slovene and its Spoken Varieties.

Master's Thesis, Utrecht University.

Ioannes Damascenus. Sacra Parallela (recensiones secundum alphabeti litteras

dispositae, quai tres libros conflant) (fragmenta e cod. Vat. gr. 1236),

PG.95.1040-1588; PG.96.9-441.

Chidambaram 179

Ishii, Toru. 2004. "The Phase Impenetrability Condition, the Vacuous Movement

Hypothesis, and That- t Effect". Lingua 114: 183-215.

Izvorski,R., Holloway-King, T., and Rudin, C. 1997. "Against li Lowering in Bulgarian".

Lingua, 102: 187-194. Elsevier Science, B.V.

Jayaseelan, K.A. 2000. "Lexical Anaphors and Pronouns in Malayalam". Lexical

Anaphors and Pronouns in Selected South Asian Languages: a Principled

Typology, 113-168. Barbara C. Lust et al, eds. Berlin: Mouton de Gruyter.

Kandybowicz, Jason. 2008. The Grammar of Repetition. Philadelphia, PA: John

Benjamins, B.V.

Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA.: The MIT Press.

Kayne, Richard. 1972. "Subject Inversion in French Interrogatives." In Generative

studies in Romance languages, 70-126. J. Casagrande and B. Saciuk eds. Rowley,

Mass.: Newbury House.

Kennedy, Christopher. 2002. "Comparative Deletion and Optimality in Syntax". Natural

Language and Linguistic Theory 20.3, 553-621.

Kennedy, Christopher. 1997. "Antecedent Contained Deletion and the Syntax of

Quantification". Linguistic Inquiry 28.4, 662-688.

Kennedy, Christopher and Merchant, Jason. 1997. "Attributive Comparatives and Bound

Ellipsis". Linguistics Research Center report LRC-97-03, University of

California, Santa Cruz.

Koopman, H. and Sportiche, D. 1982. "Variables and the Bijection Principle". Linguistic

Review 2, 139-160.

Kotzolglou, G. and Varakosta, S. 2005. "Clitics in Greek Restrictive Relatives: an

Integrated Approach." Reading Working Papers in Linguistics 8: 27-49. Reading,

UK.

Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and

Projections. Doctoral dissertation. MIT. Cambridge, MA.

Longobardi, Giuseppe. 1994. "Reference and Proper Names: A theory of N-movement in

Syntax and Logical Form". Linguistic Inquiry 25.4, 609-665.

Marušič, F. and Žaucer, R. 2006. "The definite article in colloquial Slovenian and an AP

related DP position". Formal Approaches to Slavic Linguistics 14: The Princeton

Meeting, 189-204. Lavine, J. et al, eds. Ann Arbor: Michigan Slavic Publications.

Chidambaram 180

Milsark, Gary. 1977. "Toward an explanation of certain peculiarities of the existential

construction in English". Linguistic Analysis 3: 1-29.

Mistry. P.J. 2000. "Lexical Anaphors and Pronouns in Gujarati". Lexical Anaphors and

Pronouns in Selected South Asian Languages: a Principled Typology, 333-396.

Barbara C. Lust et al., eds. Berlin: Mouton de Gruyter.

McCloskey, James. 2006. "Resumption". The Blackwell Companion to Syntax, 94–117.

Martin Everaert and Henk van Riemsdijk, eds., Oxford: Blackwell.

McCloskey, James. 2002. "Resumption, successive cyclicity, and the locality of

operations". In Derivation and Explanation in the Minimalist Program, ed.

Samuel Epstein and Daniel Seeley, 184–226. Oxford, Blackwell Publishers.

McCloskey, James. 1991. "Clause structure, ellipsis and proper government in Irish".

Lingua 85:259–302.

Mitrović, Ivana. 2011. "The case of Nominative in Serbo-Croatian relative clauses".

Paper presented at Formal Approaches to Slavic Linguistics 20: the MIT Meeting.

May 2011. Cambridge, MA.

Montalbetti, Mario. 1984. After Binding. Unpublished doctoral dissertation. MIT.

Cambridge, MA.

Newman. Paul. 2000. The Hausa Language: An Encyclopedic Reference Grammar. New

Haven and London: Yale University Press.

Panagiotidis, Phoevos. 2002. Pronouns, Clitics, and Empty Nouns: 'Pronominality' and

Licensing in Syntax. Amsterdam, the Netherlands: John Benjamins.

Panagiotidis, Phoevos. 2000. "Demonstrative determiners and operators: the case of

Greek". Lingua 110, 717-742

Perlmutter, David. 1971. Deep and Surface Structure Constraints in Syntax. New York:

Holt, Rinehart, and Winston.

Pesetsky, David. 2007. "Undermerge... and the secret genitive inside every Russian

noun".Paper presented at Formal Approaches to Slavic Linguistics 16: The Stony

Brook Meeting. May 2007, Stony Brook, NY.

Pesetsky, David & Esther Torrego. 2001. "T-to-C Movement: Causes and Consequences".

Ken Hale: A Life in Language, 355-426. Michael Kenstowicz, ed., Cambridge,

MA: MIT Press.

Chidambaram 181

Pesetsky, David. 1998. "Some optimality principles of sentence pronunciation". Is the

best good enough?, 337-383. Barbosa, P., Fox, D., McGinnis, M., Hagstrom, P.,

and Pesetsky, D., eds. Cambridge, MA: MITWPL and the MIT Press.

Pesetsky, David. 1982. "Complementizer-trace Phenomena and the Nominative Island

Condition". The Linguistic Review 1: 297-343.

Richards, Norvin. 2002. "A Distinctness Condition on Linearization". Proceedings of

West Coast Conference on Formal Linguistics 20, 470-483. K. Megerdoomian

and L. Bar-el, eds.. Somerville, MA: Cascadilla Press.

Ross, John R. 1967. Constraints on Variables in Syntax. Doctoral dissertation, MIT.

Cambridge, MA.

Rowling, J.K. 1999. Harry Potter and the Prisoner of Azakaban. London, U.K.:

Bloomsbury. Translated by Jakob J. Kenda as Harry Potter in Jetnik iz Azkabana.

(Ljubljana, Slovenia: EPTA. 2003.)

Safir, Ken. 1998. "Reconstruction and Bound Anaphora: Copy Theory without deletion at

LF". ms., New Brunswick, NJ: Rutgers University.

Sauerland, Uli. 2003. "Unpronounced Heads in Relative Clauses". The Interfaces:

Deriving and Interpreting Omitted Structures, 205-226. Kerstin Schwabe and

Susanne Winkler eds. Amsterdam: John Benjamins.

Sauerland, Uli. 1998. The meaning of chains. Doctoral dissertation, MIT, Cambridge,

MA.

Sells, Peter.1984. Syntax and Semantics of Resumptive Pronouns. Doctoral dissertation,

University of Massachusetts, Amherst.

Siewierska, Anna. 2004. Person. Cambridge, UK.: Cambridge University Press.

Simpson, Andrew. 2000. Wh-Movement and the Theory of Feature-Checking.

Philadelphia, PA and Amsterdam, the Netherlands: John Benjamins.

Shlonsky, Ur. 1992. "Resumptive Pronouns as a last resort". Linguistic Inquiry 23, 3:

443-68.

Sharvit, Yael. 1999. "Resumptive Pronouns in Relative Clauses". Natural Language and

Linguistic Theory 17, 3: 587-612. Kluwer Academic Publishers. the Netherlands.

Sportiche, Dominique. 1992. "Clitic Constructions". ms. University of California, Los

Angeles.

Chidambaram 182

Sportiche, Dominique. 1988. "A Theory of Floating Quantifiers and its corollaries for

constituent structure". Linguistic Inquiry 19:425-49.

Stavrou, M. and Horrocks, G. 1989. "Enklitikes ke dhiktikes antonimies mesa stin OF"

["Enclitic and Demonstrative Pronouns inside NP"]. Proceedings of the 10th

Annual Meeting of the Department of Linguistics, Faculty of Philosophy, 225-

243. Thessaloniki: Aristotle University of Thessaloniki.

Stowell, Timothy. 1981. "Complementizers and the Empty Category Principle."

Proceedings of the Eleventh Annual Meeting of the North Eastern Linguistic

Society. pp. 345-363. V.Burke and J. Pustejovsky, eds. Amherst, Mass.: GLSA.

Takahashi, Shoichi and Fox, Danny. 2005. "MaxElide and the Re-binding Problem".

Proceedings of Semantics and Linguistic Theory 15. pp. 223-240. Effi Georgala

and Jonathan Howell, eds. Ithaca, N.Y.: CLC Publications.

Torrego, Esther. 1986. "Determiners and Pronouns: a DP analysis of noun phrases in

Spanish". ms., University of Massachusetts, Boston.

Uriagereka, Juan. 1988. On Government. Doctoral dissertation, University of

Connecticut, Storrs.

Williams, Edwin. 1994. Thematic Structure in Syntax. Cambridge, MA: the MIT Press.

Williams, Edwin. 1978. "Across-the-board Rule Application". Linguistic Inquiry 9, 31-

43.

Zaenan, A., Engdahl, E., and Maling, J. 1891. "Resumptive Pronouns can be

Syntactically Bound". Linguistic Inquiry 12, 679-682.